Quantcast

Page 1 of 56 123451151 ... LastLast
Results 1 to 15 of 840

Thread: Review of Retro Gamer's 32X "Retroinspection"

  1. #1
    I remain nonsequitur Shining Hero sheath's Avatar
    Join Date
    Jul 2010
    Location
    Texas
    Age
    42
    Posts
    13,313
    Rep Power
    130

    Default Review of Retro Gamer's 32X "Retroinspection"

    I just finished my fourth draft review of Damien McFarren's "Retroinspection: Sega 32X" as printed in Retro Gamer Issue 77. Some of the wording may change for clarification, but the content should not. As usual, I invite serious critique on any point but will insist that facts back anything up.

    See the text below or by clicking the link above the footnotes will work as well.

    Introducing his editorial on the 32-bit add-on for the Sega Genesis, Damien McFerran, co-maintainer of the Mean Machines Archive[fn]"The Mean Machine's Archive -90's Gaming Magazine," (accessed September 16, 2010) available from http://www.meanmachinesmag.co.uk/ (archive.org August 22, 2008). [/fn] and Nintendo Life[fn]"Nintendo Life," (accessed September 16, 2010) available from http://www.nintendolife.com/ (archive.org August 22, 2008). [/fn], offers his very subjective opinion. "How do you take half a decade's worth of critical and commercial success and flush it down the toilet? Easy: you release a device like the Sega 32X." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 45.[/fn] Asserting that Sega was "at the height of its powers" by 1990, but "took far too many risks" with hardware by 1995, McFerran fails to cite which industry experts claim that the "commercially disappointing" Mega CD was a prelude to the "disastrous retail performance" of the 32X.[fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 45.[/fn] McFerran similarly offers a crude history of the 32X and Saturn at retail before introducing Scot Bayless, Senior Producer at Sega of America from 1990-1994.[fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 45-49.[/fn] The subsequent several pages are exemplary of the hostility towards Sega hardware that has accumulated over the last fifteen years. The comments of Scot Bayless and Marty Franz published in these pages challenge and disprove a number of common fallacies surrounding the 32X, and even some for the Saturn's development. McFerran, however, only posits hyperbolic questions and maintains an extremely negative tone throughout his six page editorial.

    Bayless recalls the 32X development origins at Winter CES 1994 in Las Vegas. Joe Miller, Head of Sega of America Research and Development, called Bayless and a couple of others into a teleconference with Hayao Nakayama. The call resulted in a move away from an enhanced Megadrive, apparently being worked on by Sega of Japan's Hideki Sato, toward a response to Atari's Jaguar. To which Joe Miller, according to Bayless, pushed for letting Sega of America develop a solution. Nakayama agreed, and the 32X was conceived on a hotel notepad when Marty Franz sketched two Hitachi SH-2 processors with individual frame buffers. Bayless continued by explaining that Sato's "Jupiter" was squeezed out due to cost factors by the 32X. McFerran then interjects his first positive statement: "From an engineering standpoint, the machine certainly had a lot of potential." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 45.[/fn] Bayless opines that the 32X's graphics subsystem was "brilliantly simple: something of a coder's dream" with "twice the depth per pixel of anything else out there." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 45.[/fn] The way the 32X handles 3D, Bayless supposes, was not being attempted "outside the workstation market." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 45.[/fn] In spite of all of this, McFerran bookends the first and second pages with an enlarged self-quote of the 32X being "the straw that broke the camel's back" and one of Bayless more baseless assertions that 32X's mere existence made Sega "look greedy and dumb to consumers." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 45-46.[/fn]

    McFerran describes "much brow-forrowing" at Sega of Japan due to the Megadrive's third place status in Japan and their "consensus" that all available resources should have been focused on Saturn.[fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] Apparently in direct conflict with this consensus was Nakayama himself, who took notice of the Genesis' "incredible commercial performance" and Sega of America's subsequent insistence of its continued marketability. Bayless recalls "there was a consensus at Sega of America that making an add-on for the Megadrive was the right move." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] Nakayama's ensuing call to get the 32X out by the end of 1994 was due, Bayless suspects, to his doubt on whether "the Saturn would make it to market within 1994." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn]

    Suppositions about how, or whether, the 32X and Saturn would coexist seemed to abound at Sega of America at least. Despite McFerran's assertion of a consensus against the 32X in the parent company, Bayless remembers "the guys at Sega of Japan were great - especially Sato's team." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] Any perceived tension between Sega's branches at the time are attributed to the "super double-secret-crunch-mode" and frayed nerves over delays such as the two week pause when the 32X development kit had one of its SH-2s fail. "The guys in Japan were awesome," Bayless persists, "they worked their tales off to help us." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] The only problem Bayless notes was the difficulty in translating English hybrid "technical Japanese" which was being sent to Sega of America as they were being written.

    McFerran, here, fails to notice that Bayless' account shows Sega of Japan was engineering the 32X based on Sega of America's work. The writer does note that the 32X and Saturn shared the same CPUs but used them very differently. Oddly, Bayless describes the Saturn's usage of quadrilaterals as "essentially a 2D system with the ability to move four corners of a sprite ... [to] simulate 3D space." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] Continuing that the Saturn's hardware rendering was bottlenecked by a "very high" pixel overwrite rate and "memory access stalls," Bayless shows preference for his team's design.[fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] These supposed problems with the Saturn being as much dependent on the quality of the code as the 32X's ability to display 3D at all seems to escape Bayless' notice.

    The 32X doing "everything in software" but having "two fast RISC chips tied to great big frame buffers and complete control to the programmer" causes Bayless to lament that the Saturn had not adopted a similar approach. Bayless' next statement should not be missed, "that ship had already sailed long before the green light call from Nakayama." Depending on how long before January 1994 the Saturn's hardware was solidified, Bayless' recollection stands against the popular theory of any redesign of the Saturn after the first Playstation's specifications were announced in mid 1993.

    McFerran then calls for a consideration of "the state of the market at the start of 1994." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] The editor claims the 3DO and Jaguar caused "nervous glances ... from Sega and Nintendo" before asserting "16-bit games were beginning to look look terribly outdated." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] The only observation McFerran offers on the actual state of the market in 1994 is that "something was certainly needed to keep the momentum going." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] In reality the US game industry, which Sega as a company had become dependent on, had entered "a three-year slump in 1993" with Nintendo reporting twenty-four and thirty-two percent drops in profits the next two years respectively.[fn]"Nintendo Expects." Television Digest 1994, 13.[/fn] According to Financial World's Kathleen Morrison though, Nintendo actually lost forty percent of its profits from 1992 to 1993 whereas Sega's earnings dropped sixty-four percent the same year.[fn]Morris, Kathleen. "Nightmare in the Fun House. (Cover Story)." Financial World 164, no. 5 (1995): 32.[/fn] Citing NPD group data, along with some of the same articles above, journalist Steven Kent observes in his book The Ultimate History of Video Games that the US console market dropped by thirty-six percent in total revenue from 1993 to 1995. [fn]Kent, Steven L. The Ultimate History of Video Games. New York: Three Rivers Press, 2001, 500.[/fn]

    The Wall Street Journal's Jeffrey Trachtenberg observed in September of 1994 that Nintendo and Sega were preparing massive marketing budgets to battle for revenue that holiday season. The article features Sega's William White and Nintendo's Peter Main extolling their respective marketing and advertising plans before citing Michael Wallace, an analyst with UBS Securities. Wallace reported that the world wide industry was expected to "slip to $2.8 billion this year from $3 billion in 1993." [fn]Trachtenberg, Jeffrey A. "Advertising: Sega and Nintendo Prepare to Do Battle over Holiday Season Sales." Wall Street Journal - Eastern Edition (1994).[/fn] Wallace was also directly quoted as saying: "The market is saturated, because almost everyone who wants a video game has one," to which Trachtenberg summarized that Wallace considered 35 million "16-bit players" saturation for the US market. In response the article cited William White again, who stated that Sega expected the 32X to be on the market for three to five years and sell one million units by April 1995.[fn]Trachtenberg, Jeffrey A. "Advertising: Sega and Nintendo Prepare to Do Battle over Holiday Season Sales." Wall Street Journal - Eastern Edition (1994).[/fn]

    Neither McFerran or Bayless make any observations about the 32X's role in correcting the fact that Sega's revenue had declined much more sharply than Nintendo's had at the beginning of the second full year of market decline. McFerran also placed a list of "specifications" for the 32X in a sub column, which he ends by stating flatly "it would be easier to list the reasons why the system was a failure, but there are some positive things to mention." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn] The editor then reluctantly admits to the 32X launch receiving three well adapted arcade conversions, and the add-on's affordability, before tossing out that "it's just a shame that it sold so poorly because the potential was there for greatness." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 46.[/fn]

    Turning the page, Bayless notes that Sega of America was making Saturn games even while "the 32X had to ramp up ... to hit its timing." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 47.[/fn] Apparently the two system's sharing SH-2's resulted in the 32X facing shortages in the face of the "decoupled" Saturn. McFerran chimes in that Bayless' team was "essentially attempting the impossible," citing only the ten month development goal and resource management as evidence. Also, according to McFerran, the Saturn's November 1994 launch in Japan was "a bombshell that essentially wrecked the 32X's chances of any kind of success." After reflecting on the fact that Saturn and 32X were releasing much more closely than was assumed, Bayless recalls consumers started to ask "why should I buy a 32X when the Saturn is only a few Months away?" [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 47.[/fn] Yet the U.S. launch date of the Saturn was unknown by anybody, least of all consumers up to and following its surprise launch in May of 1995. So, the only framework for such a theoretical consumer question is that they might have preferred to wait ten to twelve months for the Saturn, which 1995 sales of the PS1 and Saturn significantly discredits.

    If consumers were thinking about the far off Saturn in 1994 it is not evidenced in the 32X's sales figures. McFerran also calls the early Saturn launch in Japan the thing that transformed "the 32X from a life-saving blood transfusion ... into a poisonous tumour that would further erode the company's standing in the global marketplace." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 47.[/fn] Begrudgingly admitting that "the 32X experienced a reasonably successful launch in the West" while calling the $159 launch price a "hefty price tag," McFerran reports that "the machine shifted its initial shipment of 600,000 units with ease; it was even reported at the time that demand had far outstripped supply." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 47.[/fn] Even while admitting that the same successful launch occurred in Europe McFerran only credits that to Sega's "power" in that region.

    Bayless also observes that the marketing stance of the 32X being "a bridge from Mega Drive to Saturn ... made us look greedy and dumb to consumers, something that a year earlier I couldn't have imagined people thinking about us. We were the cool kids." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 47.[/fn] Some further research needs to be done on how Sega's reputation as "the cool kids" was attained and how much that reputation had to do with the perception of cutting edge hardware. In a sub-column roughly one-third of the page is a direct print interview between Retro Gamer and Marty Franz, Vice President of technology for Sega of America from 1993-1997. Franz describes the initial call from Nakayama as "more like a summons" for Joe Miller and Steve Payne to hear Nakayama's vision in Japan. Franz recalls that "within a few weeks a small group of us [were] on a plane to Japan to discuss the product." Franz concurs with Bayless that the 16-bit market was going to earn Sega more income than the Saturn would in the near future. Adding much needed detail to the early hotel room 32X hardware sketch, Franz asserts "we really liked the Hitachi SH-2 CPUs that the Saturn had and felt they were the star of the show." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 47.[/fn]

    Much the same as Bayless' recollection of the early Saturn architecture, Franz thus corroborates that the dual SH-2s within the Saturn were already well known by the beginning of 1994. Franz also disavows any known friction between Sega of America and Sega of Japan at the time. Describing the 32X as "a fast paced development process with a great product produced at the end," Franz and Bayless agree that the two divisions of Sega worked as a team despite "a great deal of distance between us." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 47.[/fn]

    Only the dual SH-2s were in common between the 32X and the Saturn. The focus, according to Franz, was on the development timeline while pushing "everything we could imagine that would enable great games." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 47.[/fn] Franz concludes that the team knew that CD-ROM based next generation consoles were eventually going to take over, but thinks that any perceived deficiencies in the 32X's library were due to the system's short development cycle and lack of third generation software.

    McFerran continues on the next page by claiming that the 32X failed due to "a distinct lack of compelling software." The editor then cites Bayless, who supposes that demand fell because "the launch mix for the 32X was horrible ... some of the games were pretty good, but in context they needed to be amazing." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 48.[/fn] Bayless persists that the Playstation launch in Japan made "any argument in favor of the 32X ... ridiculous." While McFerran remains fixated on proving a design flaw, Bayless insists "the games in the queue were effectively jammed into a box as fast as possible ... those games were deliberately conservative, they did nothing to show off what the hardware was capable of." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 48.[/fn] Clearly missing the point, McFerran responds by asking whether he ever believed in the 32X, to which Bayless replies "32X was a great hypothesis ... but in execution it was disastrous." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 48.[/fn] The rest of Bayless analysis had to do with "hardware ... industrial design and games" being "late or buggy - or both." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 48.[/fn] "I spent weeks working with id Software's John Carmack," Bayless continues "who literally camped out at the Sega of America building in Redood City trying to get Doom ported ... and he still had to cut a third of the levels." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 48.[/fn] Whether the time frame caused the 32X version of Doom to be limited to 3 Megabytes, compared to the 4 Megabyte source material, is not addressed.

    The rest of the article is prefaced with another sub-panel with the Genesis Sega CD combo unit, the CDX, titled "Firing Blanks" and the Megadrive, Mega CD and 32X dubbed "Three's a crowd." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 49.[/fn] Bayless continues his previous thoughts by claiming that Sega should have dropped the 32X in the Spring of 1994. By 1995, Bayless recalls, "everybody knew ... admitting it publicly that the 32X was a mistake wasn't an option." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 49.[/fn] From there things just get ugly. Bayless asserts, and McFerran makes sure to emphasize in huge quotes, "We stormed the hill and when we got to the top we realized it was the wrong damn hill." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 49.[/fn] Claiming that working for Sega from 1992 through 1994 was "like watching the Hindenburg in slow motion." Bayless goes on to list the Mega CD's "cheap consumer drives" and emphasis on Full Motion Video, "positioning the 32X as an orphan system," claiming the Saturn was just a 2D system again, and then trashing peripherals like the Activator and "Sega VR." In response to which, Bayless feels consumers "voted with their wallets and stayed away." [fn]McFerran, Damien. "Retroinspection: Sega 32x." no. 77: 49.[/fn]

    Bayless concludes by correctly explaining that the success of an organization is very much contingent on the transparency in communication in each individual group. Whether Microsoft exemplifies this motif, as Bayless claims, is certainly debatable, but clear communication within an organization is tantamount to success. Especially in light of the interviews conducted by Retro Gamer here, the 32X itself was clearly not a cause of Sega's ultimate downfall as a hardware manufacturer in 2001. McFerran observes Bayless' opinion of the 32X being a "warning" to present hardware manufacturers, but the cause of why it failed has yet to be thoroughly and objectively explored, least of all by journalists.
    Last edited by sheath; 10-13-2010 at 12:47 PM. Reason: Updated for clarity, again.

  2. #2
    Road Rasher oldmanwinters's Avatar
    Join Date
    Dec 2008
    Location
    Louisville, KY
    Posts
    329
    Rep Power
    13

    Default

    I love seeing retrospectives on that little black mushroom...

    Thanks for the sneak peak!

  3. #3
    Hero of Algol kool kitty89's Avatar
    Join Date
    Mar 2009
    Location
    San Jose, CA
    Age
    30
    Posts
    9,724
    Rep Power
    63

    Default

    Quote Originally Posted by sheath View Post
    Bayless opines that the 32X's graphics subsystem was "brilliantly simple: something of a coder's dream" with "twice the depth per pixel of anything else out there." 7 The way the 32X handles 3D, Bayless supposes, was not being attempted "outside the workstation market." 8 In spite of all of this, McFerran bookends the first and second pages with an enlarged self-quote of the 32X being "the straw that broke the camel's back" and one of Bayless more baseless assertions that 32X's mere existence made Sega "look greedy and dumb to consumers." 9
    The color depth thing is purely false: the Jaguar and 3DO both had higher color depths (24-bit vs 15-bit) and regularly operated at 16bpp vs the 32x's almost exclusive use of 8bpp 256 color indexed rendering. (not to mention the variable resolution of the Jag and 3DO vs fixed resolution of the 32x)

    Simple yes, but not hardly a dream... you had 2 fairly small (fine for 256 color but limited for 16bpp) and slow framebuffers that limited performance, especially due to the low bandwidth interface to the SH2s. (slow 16-bit DRAM at 1/3 the speed of the SH2s and 1/2 the bit width) And no hardware acceleration to speak of really beyond simple line fill, not even line by line affine rendering. (let alone proper blitter functionality like the Sega CD had)

    So certainly simple, but by no means a dream other than pure flexibility -and heavy dependency on low-level optimaztion with CPU grunt. (which would have been there with hardware acceleration too as the only flexibility came from the CPU resource... the simplicity of the VDP was mainly advantageous in short development cycle)

    Bayless recalls "there was a consensus at Sega of America that making an add-on for the Megadrive was the right move." 11
    The right move by SoA's choice, or the best option in the context of having to have a 32-bit system out by fall of '94? It would be really interesting to hear an account of thoughts on sticking with the Genesis and CD alone.

    Nakayama's ensuing call to get the 32X out by the end of 1994 was due, Bayless suspects, to his doubt on whether "the Saturn would make it to market within 1994."
    Interesting, so there WAS significant design work still going on with the Saturn. Granted, there's a lot of general design work in terms of fixing bugs in a design and getting it production quality, so that doesn't mean anything one way or the other. (albeit you'd think efforts would be focused on achieving that deadline rather than diverting resources to a parallel project)

    Bayless remembers "the guys at Sega of Japan were great - especially Sato's team." 13 Any perceived tension between Sega's branches at the time are attributed to the "super double-secret-crunch-mode" and frayed nerves over delays such as the two week pause when the 32X development kit had one of its SH-2s fail. "The guys in Japan were awesome," Bayless persists, "they worked their tales off to help us." 14 The only problem Bayless notes was the difficulty in translating English hybrid "technical Japanese" which was being sent to Sega of America as they were being written.
    Very interesting, though that doesn't address the feeling of upper management on the topic.

    Oddly, Bayless describes the Saturn's usage of quadrilaterals as "essentially a 2D system with the ability to move four corners of a sprite ... [to] simulate 3D space."
    That is quite true, it's an adaptation of the sprite engine and actually somewhat of an outgrowth of the mechanism used in the Sega CD's blitter.
    However, it is effectively similar to the general method of triangle rasterization too but with 4 points rather than 3 for each primitive and both are just as "2D" so to speak (you are after all interpreting a 3D environment to a 2D plane).
    Software rendering the 32x (DOS, etc) games had to simulate that in software rendering line by line, setting the end points for each slice of a polygon per line.

    Continuing that the Saturn's hardware rendering was bottlenecked by a "very high" pixel overwrite rate and "memory access stalls," Bayless shows preference for his team's design.16 These supposed problems with the Saturn being as much dependent on the quality of the code as the 32X's ability to display 3D at all seems to escape Bayless' notice.
    Let alone the fact that the Saturn could do everything the 32x could do but better (faster CPUs, 32-bit bus, more RAM, faster and larger framebuffers, etc) and then had hardware acceleration on top of that. (so if you did want to render just like the 32x you could and be significantly faster in many respects, but you also had a lot more hardware to work with -and also a lot more expense)

    The 32X doing "everything in software" but having "two fast RISC chips tied to great big frame buffers and complete control to the programmer" causes Bayless to lament that the Saturn had not adopted a similar approach. Bayless' next statement should not be missed, "that ship had already sailed long before the green light call from Nakayama." Depending on how long before January 1994 the Saturn's hardware was solidified, Bayless' recollection stands against the popular theory of any redesign of the Saturn after the first Playstation's specifications were announced in mid 1993.
    The framebuffers were small and slow compared to what the Saturn had (less than 1/3 the speed and 1/2 the size) to the extent that you could only have up to 320x204 for the high color mode on the 32x (completely filling the framebuffer), though 256 color stuff could fit in fine (still slow though). Hell they were the same as the Sega CD's word RAM but clocked slower. (7.67 MHz vs 12.5 MHz)

    It's like Bayless doesn't realize that the Saturn could be used just like the 32x, but be significantly faster and have the hardware acceleration to use on top of all that.

    His comments about the Saturn design are inconclusive at best though and doesn't disprove any such late 1993 modification, especially given the odd context of his other comments.

    The premise posed in the context of the "early" Saturn was use of software rendering like the 32x rendering to VDP2 planes, but with an SH1 rather than dual SH2s and that the SH2s and 3D VDP features were added late in the design to help compete. (whether that occured in direct response to the PSX or if it started earlier are other issues)
    It could have been that the VDP design was only modified moderately: like adding the 3D drawing functions to VDP1. (with VDP1 initially being more like a higher-color/faster incarnation of the CD's blitter in that context -which is more or less what it is now but with the 3D drawing modes and gouraud shading)
    It could have been in response to the 3DO's 3D capabilities for that matter.

    If the SH2s, 3D drawing modes, and Geometry DSP were all late additions to the design that would certainly explain some oddities of the console: the presence of the SH1 in the system at all (a very expensive chip to use simply for CD-ROM management), the poor documentation of the DSP, and the bugs with VDP1 translucency with 3D drawing modes.

    McFerran then calls for a consideration of "the state of the market at the start of 1994." 17 The editor claims the 3DO and Jaguar caused "nervous glances ... from Sega and Nintendo" before asserting "16-bit games were beginning to look look terribly outdated."
    3DO and PCs perhaps, but unfortunately the Jaguar had very little that put the 16 bit consoles to shame at the time with Doom probably the biggest exception. (a bunch of other games looked and sounded much like average 16-bit fare while a few others fell into the rushed tech demo category like Crescent Galaxy -Cybermorph had questionable art design and lacked music but had an actual game behind it at least)

    The only observation McFerran offers on the actual state of the market in 1994 is that "something was certainly needed to keep the momentum going." 19 In reality the US game industry, which Sega as a company had become dependent on, had entered "a three-year slump in 1993" with Nintendo reporting twenty-four and thirty-two percent drops in profits the next two years respectively.20 According to Financial World's Kathleen Morrison though, Nintendo actually lost forty percent of its profits from 1992 to 1993 whereas Sega's earnings dropped sixty-four percent the same year.21
    Hmm, all those figures are purely tied to the 16-bit home console market? (no arcades, no 8-bits, etc and only for North America)

    Nintendo's rather conventional approach with a strong push of new games seemed a lot less risky (and generally foolproof) compared to Sega's approach of releasing a 2nd add-on.

    Wallace was also directly quoted as saying: "The market is saturated, because almost everyone who wants a video game has one," to which Trachtenberg summarized that Wallace considered 35 million "16-bit players" saturation for the US market. In response the article cited William White again, who stated that Sega expected the 32X to be on the market for three to five years and sell one million units by April 1995.24
    Well that was obviously a false measure of the market given 35M was only about 80% saturation for the US market as it turned out, unless 80% is "almost everyone."

    And of course, there was tons of software sales to be had, especially from newer adapters with few games. Granted, keeping volumes up with increased software sales alone was a bit tougher. (higher priced games helps that if you can sell them like that )
    The falling prices of hardware would cause revenue to drop too though. (by 1994 buying 2 or 3 new games was the same as buying a console)

    Neither McFerran or Bayless make any observations about the 32X's role in correcting the fact that Sega's revenue had declined much more sharply than Nintendo's had at the beginning of the second full year of market decline.
    What are Kent's sources for his revenue figures BTW? (I finally got UVGH but I haven't had a chance to read it yet, let alone find those specific references -and I didn't see it in the cliffnotes)

    I also haven't found the source for the other set of revenue figures for SoA.

    Turning the page, Bayless notes that Sega of America was making Saturn games even while "the 32X had to ramp up ... to hit its timing." 27 Apparently the two system's sharing SH-2's resulted in the 32X facing shortages in the face of the "decoupled" Saturn.
    How was that possible if SoA had yet to receive documentation for the Saturn to even begin development? (let alone all the 32x software clearly taking priority in the west -with a few lower budget titles also coming from Japan while the big budgets went to Saturn games, especially Panzer Dragoon)

    McFerran chimes in that Bayless' team was "essentially attempting the impossible," citing only the ten month development goal and resource management as evidence.
    10 month development goal? The 32x was designed in close to 6 months... though the actual release didn't come until later. (hardware is going to have to be solidified for production a couple months before launch to allow for software and sufficient hardware to be produced) Plus the 32x was already being demoed at the Summer CES that June.
    In any case that really was too short to really build an efficient machine.

    Also, according to McFerran, the Saturn's November 1994 launch in Japan was "a bombshell that essentially wrecked the 32X's chances of any kind of success."
    That likely would only have hindered it due to conflicting development priorities going to the Saturn (also over the Genesis and CD), but that really didn't became a huge issue until the end of 1995 when the Saturn officially took priority over all else.
    The Saturn's May 1995 launch was most definitely a bombshell like that though. The JP launch was less of an issue for sure though, at least in as far as consumer perception.

    If consumers were thinking about the far off Saturn in 1994 it is not evidenced in the 32X's sales figures. McFerran also calls the early Saturn launch in Japan the thing that transformed "the 32X from a life-saving blood transfusion ... into a poisonous tumour that would further erode the company's standing in the global marketplace." 29 Begrudgingly admitting that "the 32X experienced a reasonably successful launch in the West" while calling the $159 launch price a "hefty price tag," McFerran reports that "the machine shifted its initial shipment of 600,000 units with ease; it was even reported at the time that demand had far outstripped supply." 30 Even while admitting that the same successful launch occurred in Europe McFerran only credits that to Sega's "power" in that region.
    I agree again on all but the sales figures: all cited seem to correspond to units shipped, not units sold to consumers, and demand from retailers vs demand from consumers are 2 very different things.

    Bayless also observes that the marketing stance of the 32X being "a bridge from Mega Drive to Saturn ... made us look greedy and dumb to consumers, something that a year earlier I couldn't have imagined people thinking about us.
    I thought the position was that the 32x wasn't a bridge for consumers, but a low-end mass market option while the Saturn was too expensive.
    It certainly made a crappy bridge in general given the total lack of compatibility with the Saturn. (a cut-down cart-based Saturn would have fit the bill of such a bridge, but the 32x was nothing of the sort)

    Only the dual SH-2s were in common between the 32X and the Saturn. The focus, according to Franz, was on the development timeline while pushing "everything we could imagine that would enable great games." 34 Franz concludes that the team knew that CD-ROM based next generation consoles were eventually going to take over, but thinks that any perceived deficiencies in the 32X's library were due to the system's short development cycle and lack of third generation software.
    The 32x was never going to compete on par with the Saturn or PSX, but it could have been OK as a low-end alternative. Ports from PCs or 5th gen consoles would be especially difficult due to lack of RAM as well as the cart based media, let alone the hardware limitations.

    McFerran continues on the next page by claiming that the 32X failed due to "a distinct lack of compelling software." The editor then cites Bayless, who supposes that demand fell because "the launch mix for the 32X was horrible ... some of the games were pretty good, but in context they needed to be amazing." 35 Bayless persists that the Playstation launch in Japan made "any argument in favor of the 32X ... ridiculous."
    That's patently false: if nothing else the 32x's software lineup was quite respectable, especially the games following that. But of course, the 32x wasn't purely meant to compete in the 5th gen market, but also among the late 4th gen market. (albeit Sega really should have focused purely on the late 4th gen market alone and continued competition against Nintendo while pushing into the high-end market against Sony with the Saturn -not going mainstream until late 1996 at least)

    But the 32x failed because Sega pulled the plug on it more than anything else... well the problem with Sega all around was really related to the Saturn and would have been problematic with or without the 32x if the May launch had been no different.

    While McFerran remains fixated on proving a design flaw, Bayless insists "the games in the queue were effectively jammed into a box as fast as possible ... those games were deliberately conservative, they did nothing to show off what the hardware was capable of."
    Most of the launch games (and late 1994 games in general) truly did show off the hardware in general, though a few moderate Genesis enhancements were less wowing. (Star Wars Arcade, Virtua Racing, Metal Head, Space Harrier, Cosmic Carnage, Doom, etc) Some were conservative in some respects, but all showed off things not possible on any 16 bit console. (at least at the color depth and resolution used... albeit you could certainly ague that reasonable approximations could have been done in some cases)
    I think Virtua Racing might have even used the highcolor mode, but I'm not sure. (if it didn't then no game ever did)


    The rest of Bayless analysis had to do with "hardware ... industrial design and games" being "late or buggy - or both." 38 "I spent weeks working with id Software's John Carmack," Bayless continues "who literally camped out at the Sega of America building in Redood City trying to get Doom ported ... and he still had to cut a third of the levels." 39 Whether the time frame caused the 32X version of Doom to be limited to 3 Megabytes, compared to the 4 Megabyte source material, is not addressed.
    Huh, interesting: I hand't fully realized the extent of Carmak's involvement... I assumed that that was the main reason the Jaguar port was so much more optimized in spite of being very rushed like the 32x release. (granted the 32x had to work with far less RAM and slow ROM, but you otherwise had a rather straightforward machine to work with compared to the Jaguar's bugs and bottlenecks -granted you had full hardware scaled sprites at high speed and the powerful custom RISC GPU but bugs, no cache, and a unified bus that was horribly bottlenecked by using the 68k and DSP to a lesser extent -vs the Genesis 68k with dedicated 64k bus -it had a lot of bandwidth but was tricky to make use of it well)


    Bayless continues his previous thoughts by claiming that Sega should have dropped the 32X in the Spring of 1994. By 1995, Bayless recalls, "everybody knew ... admitting it publicly that the 32X was a mistake wasn't an option." 41 From there things just get ugly.
    True, if they were going to drop it it had to be before it was publically anounce and preferably before dumping a bunch of R&D in.

    BY 1995 they were locked in and thus should have made the best of it for the long haul. The worst possible thing was cutting and running without even a decent phase out period: but that's what happened.
    6 days older than SEGA Genesis
    -------------
    Quote Originally Posted by evilevoix View Post
    Dude it’s the bios that marries the 16 bit and the 8 bit that makes it 24 bit. If SNK released their double speed bios revision SNK would have had the world’s first 48 bit machine, IDK how you keep ignoring this.
    Quote Originally Posted by evilevoix View Post
    the PCE, that system has no extra silicone for music, how many resources are used to make music and it has less sprites than the MD on screen at once but a larger sprite area?

  4. #4
    Road Rasher
    Join Date
    Jan 2009
    Location
    Green Bay, WI
    Posts
    289
    Rep Power
    13

    Default

    Good analysis sheath. This still hasn't popped up at my local B&N so I'm starting to doubt I will actually get to own it.

    Kool Kitty, you should start a blog (that's a compliment, not a snide remark).
    4 classic video game reviews every week - IMPLANTgames on YouTube

  5. #5
    I remain nonsequitur Shining Hero sheath's Avatar
    Join Date
    Jul 2010
    Location
    Texas
    Age
    42
    Posts
    13,313
    Rep Power
    130

    Default

    Thanks kgenthe, that kind of thing takes just enough time that I question the effort. Unless the material contains heretofore unpublished primary accounts I usually don't bother with that much detail and just summarize the writer's views. This was a special case.

    I think if you do manage to get the issue you'll find that most of the primary quotes are repeated here and I summarized McFerran's contribution. I really vacillated about taking a stance against McFerran's tone, but ultimately he just tried *way* too hard, and without sound reasoning, to make the entire article into a slam against the 32X and Sega. Ordinarily I'd just ignore the false or negative comments and focus on what unique info was presented, but McFerran's overview and narrative overtly glazed over the new facts.

  6. #6
    I remain nonsequitur Shining Hero sheath's Avatar
    Join Date
    Jul 2010
    Location
    Texas
    Age
    42
    Posts
    13,313
    Rep Power
    130

    Default

    Quote Originally Posted by kool kitty89 View Post
    The color depth thing is purely false: the Jaguar and 3DO both had higher color depths (24-bit vs 15-bit) and regularly operated at 16bpp vs the 32x's almost exclusive use of 8bpp 256 color indexed rendering. (not to mention the variable resolution of the Jag and 3DO vs fixed resolution of the 32x)
    Coming from the software perspective, the 32X's extensive use of Genesis or SNES source material limited its real world capabilities even further. Of course ports to the Jaguar and 3DO apparently did not receive full service as well. I own Flashback on 3DO and could just as well play the Genesis or Sega CD games instead, the differences on a television are negligible. That is, in fact, where things were with these consoles in 1994-95.

    Quote Originally Posted by kool kitty89 View Post
    Simple yes, but not hardly a dream... you had 2 fairly small (fine for 256 color but limited for 16bpp) and slow framebuffers that limited performance, especially due to the low bandwidth interface to the SH2s. (slow 16-bit DRAM at 1/3 the speed of the SH2s and 1/2 the bit width) And no hardware acceleration to speak of really beyond simple line fill, not even line by line affine rendering. (let alone proper blitter functionality like the Sega CD had)

    So certainly simple, but by no means a dream other than pure flexibility -and heavy dependency on low-level optimaztion with CPU grunt. (which would have been there with hardware acceleration too as the only flexibility came from the CPU resource... the simplicity of the VDP was mainly advantageous in short development cycle)
    To be brief, Bayless makes it clear that they tried to create a developer's dream within the short development time line. He obviously believes they succeeded, as does Franz.

    Quote Originally Posted by kool kitty89 View Post
    The right move by SoA's choice, or the best option in the context of having to have a 32-bit system out by fall of '94? It would be really interesting to hear an account of thoughts on sticking with the Genesis and CD alone.
    The context of the comment was creating an answer to the Jaguar by the end of 1994. The 32X edged out Jupiter (Sato's "bridge" design) in this effort. Bayless speculates that Nakayama had doubts about the Saturn being ready by the same time. He also makes it very clear that SoA felt the Saturn would not be in *any* market at the time of the 32X's launch.

    Quote Originally Posted by kool kitty89 View Post
    Interesting, so there WAS significant design work still going on with the Saturn. Granted, there's a lot of general design work in terms of fixing bugs in a design and getting it production quality, so that doesn't mean anything one way or the other. (albeit you'd think efforts would be focused on achieving that deadline rather than diverting resources to a parallel project)
    That's the opposite implication of what Bayless' account holds. His team felt that a fully software driven SH-2 based system was superior to the Saturn's *existing design.* The Saturn's SH-2 based design actually influenced their decision to run with the same for 32X. Now to the point, if SoJ was completely finished with the design process of the Saturn they would have launched in 1993 instead of 1994. Perhaps some of the chips were still in the final design stages. Having a static hardware design, and actually having every last detail optimized and ready for manufacturing are two different things.

    Quote Originally Posted by kool kitty89 View Post
    That is quite true, it's an adaptation of the sprite engine and actually somewhat of an outgrowth of the mechanism used in the Sega CD's blitter.
    However, it is effectively similar to the general method of triangle rasterization too but with 4 points rather than 3 for each primitive and both are just as "2D" so to speak (you are after all interpreting a 3D environment to a 2D plane).
    Software rendering the 32x (DOS, etc) games had to simulate that in software rendering line by line, setting the end points for each slice of a polygon per line.
    That's basically what I think Bayless missed, the Playstation handles 3D with 2D objects as well.

    Quote Originally Posted by kool kitty89 View Post
    Let alone the fact that the Saturn could do everything the 32x could do but better (faster CPUs, 32-bit bus, more RAM, faster and larger framebuffers, etc) and then had hardware acceleration on top of that. (so if you did want to render just like the 32x you could and be significantly faster in many respects, but you also had a lot more hardware to work with -and also a lot more expense)
    Bayless thinks that the memory timing issues created by the Saturn design, which is apparently one of the main hardware complaints, were avoided by the 32X. He does not imply that the 32X could outperform the Saturn, he states that a Saturn designed like the 32X would outperform the historical Saturn.

    Quote Originally Posted by kool kitty89 View Post
    The framebuffers were small and slow compared to what the Saturn had (less than 1/3 the speed and 1/2 the size) to the extent that you could only have up to 320x204 for the high color mode on the 32x (completely filling the framebuffer), though 256 color stuff could fit in fine (still slow though). Hell they were the same as the Sega CD's word RAM but clocked slower. (7.67 MHz vs 12.5 MHz)
    The 32X design was cost conscious and the resolution was always going to be limited to Genesis resolution. So yes, Bayless assertion of 32X having "great big frame buffers" is in the context of 320x224 and 8-bit color.

    Quote Originally Posted by kool kitty89 View Post
    It's like Bayless doesn't realize that the Saturn could be used just like the 32x, but be significantly faster and have the hardware acceleration to use on top of all that.

    His comments about the Saturn design are inconclusive at best though and doesn't disprove any such late 1993 modification, especially given the odd context of his other comments.
    Bayless' and Franz' accounts prove that the SH-2's were in the Saturn design by the end of 1994. They also prove that a totally software oriented design philosophy was already off the table by then as well.

    Quote Originally Posted by kool kitty89 View Post
    The premise posed in the context of the "early" Saturn was use of software rendering like the 32x rendering to VDP2 planes, but with an SH1 rather than dual SH2s and that the SH2s and 3D VDP features were added late in the design to help compete. (whether that occured in direct response to the PSX or if it started earlier are other issues)
    It could have been that the VDP design was only modified moderately: like adding the 3D drawing functions to VDP1. (with VDP1 initially being more like a higher-color/faster incarnation of the CD's blitter in that context -which is more or less what it is now but with the 3D drawing modes and gouraud shading)
    It could have been in response to the 3DO's 3D capabilities for that matter.
    Right, this kind of story has been around since Sony took over in 1998. I've seen a hundred variations, but the earliest ones blame all of the Saturn's complexity on the second SH-2. They are all just theories, their only usefulness is in creating questions.

    Questions like, what would a software rendered SH-1 based 3D system look like in software? Are there any facts to support the existence of an entirely SH-1 based 3D system from Sega? Could a proper combination of VDPs and helper chips create reasonable approximations of Model 1 games? I'm supposing there is no way such a solution would target Model 2 games, as even the Saturn or Playstation required significant "porting" to do that much.

    Quote Originally Posted by kool kitty89 View Post
    If the SH2s, 3D drawing modes, and Geometry DSP were all late additions to the design that would certainly explain some oddities of the console: the presence of the SH1 in the system at all (a very expensive chip to use simply for CD-ROM management), the poor documentation of the DSP, and the bugs with VDP1 translucency with 3D drawing modes.
    Sega remained focused on limiting load times, which the Saturn does quite well in the bulk of its library compared to the Playstation. My experience is that the average Saturn game's load time was around 8 seconds, while the average PS1 games were closer to 20 seconds, and at best 12 seconds. I'm fairly certain that the Sega CD included the 12.5 Mhz 68000 and 768KB of RAM for the same purpose.

    Quote Originally Posted by kool kitty89 View Post
    3DO and PCs perhaps, but unfortunately the Jaguar had very little that put the 16 bit consoles to shame at the time with Doom probably the biggest exception. (a bunch of other games looked and sounded much like average 16-bit fare while a few others fell into the rushed tech demo category like Crescent Galaxy -Cybermorph had questionable art design and lacked music but had an actual game behind it at least)
    In my opinion, the Jaguar is one of the more exemplary instances of a console which had its reputation precede it. The N64 and Playstation 2 are also on that list. Marketing was so successful that the idea that these systems were a generational leap ahead of anything else out there was propagated to the entire industry.

    Quote Originally Posted by kool kitty89 View Post
    Hmm, all those figures are purely tied to the 16-bit home console market? (no arcades, no 8-bits, etc and only for North America)
    They are all focused on the console industry first, and the majoritarians second (Sega and Nintendo).

    Quote Originally Posted by kool kitty89 View Post
    Nintendo's rather conventional approach with a strong push of new games seemed a lot less risky (and generally foolproof) compared to Sega's approach of releasing a 2nd add-on.
    Nintendo had the luxury of continued software sales dominance. Sega had a long history of creating gobs of well reviewed and polished software that the public just would_not bite in massive numbers. The two companies took to their strengths (in sales) in 1994.

    Quote Originally Posted by kool kitty89 View Post
    Well that was obviously a false measure of the market given 35M was only about 80% saturation for the US market as it turned out, unless 80% is "almost everyone."
    He was specifically talking about "16-bit Players" with that comment. Also, nobody knows in real time exactly how much more something will grow, and the market was presently in decline.

    Quote Originally Posted by kool kitty89 View Post
    What are Kent's sources for his revenue figures BTW? (I finally got UVGH but I haven't had a chance to read it yet, let alone find those specific references -and I didn't see it in the cliffnotes)

    I also haven't found the source for the other set of revenue figures for SoA.
    I have all of them in the footnotes on my page. The quotes Kent uses were quoted in their entirety the first time I brought them up, but the section is at the end of "The Next Generation: Part One" chapter.

    Quote Originally Posted by kool kitty89 View Post
    How was that possible if SoA had yet to receive documentation for the Saturn to even begin development? (let alone all the 32x software clearly taking priority in the west -with a few lower budget titles also coming from Japan while the big budgets went to Saturn games, especially Panzer Dragoon)
    The account is that SoA was developing Saturn software during the 32X development process. It does not say how much software or how much documentation they had received on the Saturn. The 32X documentation was being sent from SoJ to SoA on the fly though. It would seem reasonable that the Saturn's docs were being relayed in similar fashion as they were being written.

    Quote Originally Posted by kool kitty89 View Post
    10 month development goal? The 32x was designed in close to 6 months... though the actual release didn't come until later. (hardware is going to have to be solidified for production a couple months before launch to allow for software and sufficient hardware to be produced) Plus the 32x was already being demoed at the Summer CES that June.
    In any case that really was too short to really build an efficient machine.
    They had to have it with games on the market in 10 months. Actually, the mandate from Nakayama might have allowed a December release date, but the team managed better.

    Quote Originally Posted by kool kitty89 View Post
    That likely would only have hindered it due to conflicting development priorities going to the Saturn (also over the Genesis and CD), but that really didn't became a huge issue until the end of 1995 when the Saturn officially took priority over all else.
    The Saturn's May 1995 launch was most definitely a bombshell like that though. The JP launch was less of an issue for sure though, at least in as far as consumer perception.
    Since Bayless left the company at the end of 1994, the evidence suggests that the Saturn's Japanese launch damaged the culture at Sega of America.

    Quote Originally Posted by kool kitty89 View Post
    I agree again on all but the sales figures: all cited seem to correspond to units shipped, not units sold to consumers, and demand from retailers vs demand from consumers are 2 very different things.
    Using sales to retailers was an "innovation" of Sony, not Sega. Sega always focused on sell through to consumers. This would be an oddity if the 600k figure in the US that McFerran claims but does not cite was sales to retailers. I also never saw stacks of 32X boxes anywhere, and Kent's book goes into some detail at the lukewarm reception by retailers in general.

    Quote Originally Posted by kool kitty89 View Post
    I thought the position was that the 32x wasn't a bridge for consumers, but a low-end mass market option while the Saturn was too expensive.
    It certainly made a crappy bridge in general given the total lack of compatibility with the Saturn. (a cut-down cart-based Saturn would have fit the bill of such a bridge, but the 32x was nothing of the sort)
    Not a bridge to the Saturn, a bridge to the 32-bit generation. The cost effectiveness of the add-on was roughly similar to buying a Super Nintendo and would have consistently brought "32-bit" gameplay if given more time.

    Quote Originally Posted by kool kitty89 View Post
    The 32x was never going to compete on par with the Saturn or PSX, but it could have been OK as a low-end alternative. Ports from PCs or 5th gen consoles would be especially difficult due to lack of RAM as well as the cart based media, let alone the hardware limitations.
    The 32X' sales alone were never going to surpass the weekly sales of Genesis consoles in general. Whether or not these exceeded Saturn or PS1 sales in mid to late 1995 is basically irrelevant. As the interviews indicate, they only expected it to last for three years maximum. As for how well they would have handled adaptations of Saturn or PS1 games, that's something we will likely never know. Though Spear 3D running on the 32X like it does, and the other tech demos we have seen, shows a lot of promise for the time.

    Quote Originally Posted by kool kitty89 View Post
    That's patently false: if nothing else the 32x's software lineup was quite respectable, especially the games following that. But of course, the 32x wasn't purely meant to compete in the 5th gen market, but also among the late 4th gen market. (albeit Sega really should have focused purely on the late 4th gen market alone and continued competition against Nintendo while pushing into the high-end market against Sony with the Saturn -not going mainstream until late 1996 at least)
    I think Bayless is coming from the perspective of what he knew the 32X was designed to do and comparing that to the launch games. In that respect he's right, his phraseology just pisses all over the quality of the library though.

    Quote Originally Posted by kool kitty89 View Post
    But the 32x failed because Sega pulled the plug on it more than anything else... well the problem with Sega all around was really related to the Saturn and would have been problematic with or without the 32x if the May launch had been no different.
    I don't think this is an either or thing, I think it's a revenue thing. Sega was losing the Arcade buffer, and that was what caused all of these relatively spontaneous shifts in the way the company had been running.

    Quote Originally Posted by kool kitty89 View Post
    Most of the launch games (and late 1994 games in general) truly did show off the hardware in general, though a few moderate Genesis enhancements were less wowing. (Star Wars Arcade, Virtua Racing, Metal Head, Space Harrier, Cosmic Carnage, Doom, etc) Some were conservative in some respects, but all showed off things not possible on any 16 bit console. (at least at the color depth and resolution used... albeit you could certainly ague that reasonable approximations could have been done in some cases)
    I think Virtua Racing might have even used the highcolor mode, but I'm not sure. (if it didn't then no game ever did)
    Take a screenshot from an emulator and take a color count, it has been a while, but I don't think Virtua Racing is hitting 256 colors. Metal Head has framerate issues when the textures get complex, Doom was cut down, Space Harrier is great and to heck with anybody who says otherwise, and Cosmic Carnage is a misunderstood stroke of genius that wasn't completed.

    I don't own any of the EA ports, and only own one of the Akklaim games I think, but I find the rest of the library more compelling than most of the other games that came out at the same time.

    Quote Originally Posted by kool kitty89 View Post
    Huh, interesting: I hand't fully realized the extent of Carmak's involvement... I assumed that that was the main reason the Jaguar port was so much more optimized in spite of being very rushed like the 32x release. (granted the 32x had to work with far less RAM and slow ROM, but you otherwise had a rather straightforward machine to work with compared to the Jaguar's bugs and bottlenecks -granted you had full hardware scaled sprites at high speed and the powerful custom RISC GPU but bugs, no cache, and a unified bus that was horribly bottlenecked by using the 68k and DSP to a lesser extent -vs the Genesis 68k with dedicated 64k bus -it had a lot of bandwidth but was tricky to make use of it well)
    I don't think Carmack's degree of involvement was known previously. It will be interesting (and great fun) to see what Chilly can come up with eventually on the Doom front.

    Quote Originally Posted by kool kitty89 View Post
    True, if they were going to drop it it had to be before it was publically anounce and preferably before dumping a bunch of R&D in.

    BY 1995 they were locked in and thus should have made the best of it for the long haul. The worst possible thing was cutting and running without even a decent phase out period: but that's what happened.
    I agree, the time to cancel the 32X was before it was announced not after. Still, I don't think Sega as an organization had a better solution for 1994's revenue problems. It makes me sick that the Genesis and Sega CD software libraries alone didn't create an upswing.
    Last edited by sheath; 10-04-2010 at 08:55 AM.

  7. #7
    Hero of Algol kool kitty89's Avatar
    Join Date
    Mar 2009
    Location
    San Jose, CA
    Age
    30
    Posts
    9,724
    Rep Power
    63

    Default

    Quote Originally Posted by sheath View Post
    Coming from the software perspective, the 32X's extensive use of Genesis or SNES source material limited its real world capabilities even further. Of course ports to the Jaguar and 3DO apparently did not receive full service as well. I own Flashback on 3DO and could just as well play the Genesis or Sega CD games instead, the differences on a television are negligible. That is, in fact, where things were with these consoles in 1994-95.
    Actually, ports from ANY other system weren't going to be very helpful for a smooth transition. PCs and newer consoles had much more RAM and mass storage making cross-platform 32x games very tough, let alone assembly language programming. (be it other games written in C or assembly in another architecture -especially x86- that's a lot of work to source a port)

    For straight arcade ports, the 32x wasn't too much worse off than contemporaries as most games required substantial redesign for ANY platform

    The best example of a multiplatform 2D game making good use of the 32x would probably be Blackthorne, but that doesn't really show off that much other than color and animation (no scrolling, not much going on on-screen etc). It is interesting to note that pretty much the entire game is 32x generated with only the Status Bar handled by the Genesis and that all the sound is also handled by the 32x's PWM with no FM or PSG. (however, by definition that means the game is not really taking full advantage of the hardware potential for making the most of the Genesis AND 32x hardware, though at least it looked a lot better than other games that were relatively simple upgrades over the genesis versions -sometime the most significant aspect was increased ROM size allowing larger samples and/or animation -things that would have made the Genesis version better had that taken priority -let alone a Sega CD release or enhanced carts)

    To be brief, Bayless makes it clear that they tried to create a developer's dream within the short development time line. He obviously believes they succeeded, as does Franz.
    I really don't see it though, but in any case: such a short time period pretty much killed any chance for reasonable hardware support, so going simple with hefty off-the-shelf CPU grunt was a good option in that respect.
    Though really, as it was there's some really simple things they could and should have put more emphasis on like adding basic blitting and affine line rendering capabilities to the VDP. (rather than adding unnecessary things like the logic for run length encoding)
    Plus, it's not like they had to work from the ground up: they could easily have built on a pre-existing design to speed things up, at least if SoJ was willing to open up such resources. (I'm really interested in what Hideki Sato had planned for the Mars project, but if that was no good: they already had the SSP-1601 licensed and the Sega CD hardware to work off of as well)

    The context of the comment was creating an answer to the Jaguar by the end of 1994. The 32X edged out Jupiter (Sato's "bridge" design) in this effort. Bayless speculates that Nakayama had doubts about the Saturn being ready by the same time. He also makes it very clear that SoA felt the Saturn would not be in *any* market at the time of the 32X's launch.
    Jupiter was a mass market name applied after the fact: Sato's mars design (AKA Super Mega Drive) was totally separate from the proposed direct Saturn derivative from everything I've seen. In fact, I've never seen any very useful info on Sato's mars design beyond Pettus's simplified quotes from Joe Miller regarding "128 colors."
    If the "Jupiter" (ie totally incompatible with MD but directly forward compatible with Saturn) design was ever on the table, it was a big mistake to drop it, but given Joe miller's comments about Sato's Mars being rather weak and barely more than a "Genesis with double the colors" seems to totally contradict that.

    I don't see where in the article it implies that SoA thought the Saturn wasn't goign to be out anywhere by 1994. It does say that Nakayma was possibly skeptical about the 1994 deadline being met, but that's hardly definitive. (unless I overlooked something)


    That's the opposite implication of what Bayless' account holds. His team felt that a fully software driven SH-2 based system was superior to the Saturn's *existing design.* The Saturn's SH-2 based design actually influenced their decision to run with the same for 32X. Now to the point, if SoJ was completely finished with the design process of the Saturn they would have launched in 1993 instead of 1994. Perhaps some of the chips were still in the final design stages. Having a static hardware design, and actually having every last detail optimized and ready for manufacturing are two different things.
    No, I meant that the Saturn was still in development during early 1994 in the context of Nakayma doubting it wouldn't be ready for a release that year.
    Having a design still yet to be completed can still have the hardware pretty much set but require a LOT of final work to eliminate bugs and actually make it production quality. (especially depending on the budget you have to work with... in the Jaguar's case the design work started in 1989, the core logic design was completed and laid down in 1990 while the next 3 years comprised actually implementing that in the real world -albeit that was a very tight and highly integrated design and more conservative and less integrated designs would have taken much less and allowed more modifications time but cost far more and/or been lower performance -in the Saturn's case, the VDP-1 definitely seems like it could have been modified relatively late)

    He has no grounds for criticizing the Saturn in any case though as it was universally more powerful and flexible than the 32x and no more difficult to program for in apples to apples comparisons. (in fact, generally EASIER due to the greater system resources)
    From a cost standpoint there's a big difference, but he wasn't arguing that.


    That's basically what I think Bayless missed, the Playstation handles 3D with 2D objects as well.
    So does the 32x and every other platform ever made that renders 3D as such. That's how polygonal 3D works, 2D primitives warped to represent 3D space using based on plotted 3D coordinates.

    The ONLY reason the Saturn and 3DO get criticized for "not being 3D enough" if the use of quadrilateral primitives going against common convention, but Sega's own Model 3 did the same thing as is praised as the greatest 3D arcade system of its time.


    Bayless thinks that the memory timing issues created by the Saturn design, which is apparently one of the main hardware complaints, were avoided by the 32X. He does not imply that the 32X could outperform the Saturn, he states that a Saturn designed like the 32X would outperform the historical Saturn.
    I'm not sure what memory timing issues that would be, but the 32x had a ton of bottlenecks to deal with in that regard. (namely slow framebuffers and ROM accessing, more so if you were to actually USE the genesis hardware -shared ROM accessing, DMA, etc)

    I'm also not sure what "like the 32x" would mean either: if you meant software based, then no, the Saturn would have been much weaker then and no games would have looked any better. (as it was software rendering was most definitely a real option, and from what I understand some developers used that to avoid the difficulties of working with quadrilaterals and used VDP1 only for texture mapping and shading, not full rasterization of polygons -such games would look far less competitive than those built around quad engines though)
    Unless you're saying that instead of ANY significant hardware acceleration, the Saturn should have been built around a CPU+framebuffer set-up with simple sound as well: in which case youd want to go with a much more powerful CPU and much more (and faster) RAM. (say an SH3 with similar framebuffers as the Saturn and 2MB of SDRAM for main/shared system RAM, but only a simple bitmap display like the 32x)

    But most such systems are more expensive than those using dedicated hardware and similar cost optimization (which the Saturn didn't emphasize no is NOT comprable). The N64 sort of did that with one R4000 based CPU and another R4000 as the GPU but with some modifications and customizations for its role. (and then added hardware integrated into it for video generation and some other things)
    If you wanted general flexibility and low cost: a custom GPU aiming at that would also be a very good move in terms of cost effectiveness, and that's what the Jaguar did (with rather limited resources at that -albeit being forced to rush out with some bugs) but even that was in addition to a capable blitter and the Object Processor in the Jaguar.
    Hell, the jaguar chipset was right up Sega's alley in terms of performance goals in general but with extreme low-cost emphasis: had Sega been the one Flare had partnered with (or Atari sold it off when still in development in 1992/93), Sega's resources could have really shaped it into something amazing and still very cost effective for 1994. (namely pushing for the remaining bugs to be worked out and replacing the 68000 with an SH2 -very likely would have worked given the flexibility of the jag's chipset- and you'd have an amazing system to work with, even more so if they'd filled the 2nd bank of DRAM for 4 MB total -they might have switched to SDRAM and doubled the bandwidth, but that would have been against the low-cost design focus) -Any such changes would have pretty much eliminated most of the programming difficulties of the Jaguar as well. (the historical Jaguar itself was VERY low cost and probably could have even dropped close to the 32x's launch price had a company like Sega been marketing it -ie higher production volumes and selling at cost vs the low volumes and profit margins Atari couldn't afford to drop any lower at the time)

    Granted, with Sega' talent, they could have likely done similar themselves had they pushed for such a design back in '90/91 and worked from there, but that's not what Sega was pushing.

    The 32X design was cost conscious and the resolution was always going to be limited to Genesis resolution. So yes, Bayless assertion of 32X having "great big frame buffers" is in the context of 320x224 and 8-bit color.
    Well, then that's a patently false and misleading statement as 320x224x256 color (from 15-bit RGB) was NOT a major thing by that point and was pretty much what mid-range PC games had been pushing for several years at that point. (and well below that of the Jaguar and 3DO) It's also not far from what the SNES had been pushing either, albeit that was using 16-color tiles and sprites with 256 indexes from 15-bit RGB and at 256x224/239.

    Bayless' and Franz' accounts prove that the SH-2's were in the Saturn design by the end of 1994. They also prove that a totally software oriented design philosophy was already off the table by then as well.
    You mean by the end of 1993, not '94 right?

    And no, while I'm sure pure software rendering using CPU resource was NEVER a serious consideration for the Saturn (and really shouldn't be for any non-rushed system) software rendering is ALWAYS an option: the Saturn could and DID do it better than the 32x, but it had a lot of added hardware to help out and that's where it shined.

    If the SH2s were added in a redesign, that likely was the FIRST change as that would be the only off-the-shelf addition and if the redesign began in late fall of 1993 (months before Mars was brought to SoA), the SH2s easily could have been part of the Design by then, even if not fully prototyped yet and even if the other hardware was also being modified.
    But again, I'm sure SoJ would never have been going an all-software oriented route as it wouldn't make sense, especially before the SH2 became an option. (SH1 would have been weak in that category and alternatives would have been expensive, especially in early/mid 1993)
    The only software rendered portion might have been 3D if VDP1 wasn't there at all, but VDP2 already is a hefty chunk of hardware acceleration with its 6 planes (and without VDP1, the SH1 would be rendering to those planes), but it's pretty likely that there was something in place of VDP1, again something very like VDP1 without the 3D drawing modes (ie a faster and higher-color successor to the Sega CD ASIC) as the 3D quad warping would build-on to such a design relatively simply. (without those warped quads, the VDP1 would be just as useful as it is for any Saturn game using software triangle rasterization -ie using the VDP only for texture mapping and shading as well as for 2D objects -rather like what the Sega CD does) VDP2 was the real monster in the Saturn though, and if you took that out and instead integrated simple video generation (like the 32x's VDP) into VDP1 and dropped VDP2's RAM it would have been significantly cheaper but much worse at 2D. (or 3D floor/ground/water/BG effects done by VDP2)

    but the earliest ones blame all of the Saturn's complexity on the second SH-2. They are all just theories, their only usefulness is in creating questions.
    That's false, the 2nd SH2 is only a small part of it: it is indeed tough to really get good performance out of the slave CPU and generally needs to be used for some specific tasks to make efficient use of the shared bus. (at best you probably get 60% performance if that) But it's added resource, they don't force you to use it.
    The main problems are not related to the CPUs at all though, but the general hardware and software development tools not being aimed at high-level programming, and due to the hardware, even with later high-level tools it was far less efficient than the PSX which was designed specifically for that like most later consoles (Dreamcast is WAY up there but GC, Xbox, and 360 are too).
    And then there's the use of quad rasterization on top of all that.
    But the generally poor early documents (even for low-level programming they were poor and omitted documentation of key features like the geometry DSP) and improved tools were late to the game. (and even if they had decent comprehensive high and low-level tools from the start it wouldn't have been nearly as easy to get good 3D performance out of as the PSX, but at least a lot better than it was)


    Questions like, what would a software rendered SH-1 based 3D system look like in software? Are there any facts to support the existence of an entirely SH-1 based 3D system from Sega? Could a proper combination of VDPs and helper chips create reasonable approximations of Model 1 games? I'm supposing there is no way such a solution would target Model 2 games, as even the Saturn or Playstation required significant "porting" to do that much.
    If it had no VDP1, SH2s, or geometry DSP, then yes pretty much limited to model 1 ports and the like (simple flat shaded polygon games, very limited texture mapping, and other software rendering like height maps/ray casting/"voxel" engines) all of which would likely be perhaps moderately close to what the 32x was outputting. (if really optimized at least, but even then probably not as good as the 32x in 3D, still awesome in 2D though, way better than the 32x+Genesis due to VDP2)
    But it was very likely there was some VDP1 for a blitter/sprite engine as I said, albeit in that context it likely had no 3D modes (for warped quads) and the DSP likely was also missing (presuming that was "added" late too, hence the poor documentation), you had a good bit more help with the addition of hardware scaled/rotated sprite/object rendering and also affine line rendering. (meaning more or less that you could have full texture mapped 3D with a similar polygon count as you would solid shaded polygons) Add that DSP and you've got a lot more to work with as it would handle all the 3D math to generate the verteces while the CPU would only have to rasterize the polygons and texture map/shade with the help of VDP1. (in which case it might look like 32x games with full texture mapping if not a bit beyond that) Likely better than the 3DO by a fair margin.

    OTOH, if you add in everything but the SH2s: full VDP1+DSP as the saturn had but with the SH1 as the CPU, the only limits to managing Saturn quality games is game logic complexity and software rendering. (ie any games that didn't kill the SH2s with game logic/AI or intensive software graphics tasks could be pretty much the same as they were in the Saturn) The problem lies in that many Saturn games (from what I understand) don't make full use of the hardware due to difficulty of porting or general poor documentation, so the slave SH2 was often used for the geometry calculations while the DSP sat idle and some games even pushed software triangle rasterization. (the real cases where the SH2s have an advantage in the Saturn is for ray-casting and such where the hardware doesn't help -a pure Doom port might have fit well on the Saturn due to that but the PSX port was redesigned as a polygon renderer due to its speed at that -and a Saturn sans the SH2s might have been best off doing a quad renderer for that too, and the 3DO definitely as its CPU was really weak by comparison and the hardware was no good at ray-casting but built around texture mapped and gouraud shading quadrilaterals and had a geometry DSP coprocessor as well)

    A fairly flexible DSP could fill the gap for needed flexibility better/cheaper than a powerful CPU though, but I'm not sure if the Saturn's Geometry DSP fits that role or not. (the PSX's GTE doesn't, but the Super FX and SVP are relatively flexible and useful for 3D math as well as rendering tasks including aiding ray-casting -or texture mapping, scaling, etc in the lack of better hardware support)
    In fact, in the premise of an earlier weaker Saturn, adding a fairly fast, cost-effective, flexible DSP coprocessor would have been a good consideration, and if the geometry DSP isn't in fact such a DSP, you could certainly argue that adding 1 SH2 and an SSP-1601 (given they had a license already) would have been better in many cases than dual SH2s and likely would have saved a fair bit of cost. (the SH2 really is a much better choice in general, but they probably could have made do with the SH1 for a lot -albeit if they went with the SH2 they REALLY should have removed the SH1 entirely to cut costs and use a simpler cheap MCU to manage the CD drive)
    You could certainly argue the same for the 32x for having an SVP+SH2 over dual SH2s, let alone a bit more meat to the VDP. (namely blitter functionality akin to the Sega CD's ASIC or even less than that)


    Sega remained focused on limiting load times, which the Saturn does quite well in the bulk of its library compared to the Playstation. My experience is that the average Saturn game's load time was around 8 seconds, while the average PS1 games were closer to 20 seconds, and at best 12 seconds. I'm fairly certain that the Sega CD included the 12.5 Mhz 68000 and 768KB of RAM for the same purpose.
    I'm not sure what the hell you're talking about? The Sega CD's added CPU and RAM have nothing to do with load times, but the 16 kB CD-ROM cache does (compared to the cacheless PCE CD), that would have been true if a cheap simple embedded MCU was used in place of the 68k.
    Likewise the Saturn had a hefty 512 kB CD-ROM cache/buffer that the SH1 worked in (but again, like the CD's CPU it was way overkill) vs the 32 kB cache of the 3DO and PSX. (that's the same reason the CD-Z loads faster than the Neo Geo CD, and I think the Wondermega/X'Eye also have enhanced caching over the standard MCD)

    There's also other things like on-the-fly data decompression to speed up loading, but that's possible on all those systems (and the PSX has a hardware decompression engine for that very purpose), and the one final way to speed up loading is putting tons of redundant data on the disc to greatly reduce seek times and allow linear loading. (but that only works if you have enough extra space on a disc or are willing to use more discs -that's also the main reason pirated Dreamcast ISOs require lots of seeking compared to GD-ROMs)

    In my opinion, the Jaguar is one of the more exemplary instances of a console which had its reputation precede it. The N64 and Playstation 2 are also on that list. Marketing was so successful that the idea that these systems were a generational leap ahead of anything else out there was propagated to the entire industry.
    But the Jaguar failed to really get mass market attention and general support, and especially lacked the funding necessary for decent media saturation. Production volumes and distribution were also critical, but the man problems were stiff 16-bit competition followed by Sony roaring in. Plus the general lack of funds for sufficient 1st party software. (had they not been strapped for cache, the Jag likely would have been held back until mid 1994 in any case)
    As it was, the 3DO outsold it by more than an order of magnitude and it really failed to reach critical mass on the mainstream marketplace. They lacked the funds and the brand name to pull it off and made other mistakes that lost their chances in Europe. (the Tramiels COULD have addressed much of that by investing tens of millions from private funds, but that was a huge risk and what they did certainly turned out well for them in any case -bringing Atari Corp out of debt in 1993 and allowing the very profitable liquidation in 1996)

    They are all focused on the console industry first, and the majoritarians second (Sega and Nintendo).
    I'm still not sure what that means as both Sega and especially Nintendo had older consoles still on the market (unless that was purely North America, in which case only Nintendo's NES mattered). On top of that it seems to imply revenue at one point and then mention drops in profits which are independent of revenue and are a different context. (competitive pricing and advertising could easily cut into profits but not necessarily revenue depending on increased sales) And then there's pure sales figures separate from revenue or profits. (another thign competitive pricing affects)

    Nintendo had the luxury of continued software sales dominance. Sega had a long history of creating gobs of well reviewed and polished software that the public just would_not bite in massive numbers. The two companies took to their strengths (in sales) in 1994.
    Sega had hits more spread out than Nintendo, but a LOT of software sales and I've not seen any comprehensive figures that pointed to Nintendo being strong in that area, especially if you exclude Japan entirely.

    If the software wasn't selling in spite of the quality there's 2 main reasons: marketing wasn't right and/or the game wasn't catering well enough to the mass market. (difficulty, gameplay mechanics, art style, etc) And then comes brand recognition, but by the mid 90s Sega had that down in North America and in most of Europe as well, albeit not the established series of Nintendo. (Mario, Metroid, Zelda, etc vs Sonic, Phantasy Star, and a few others to a lesser extent -not getting into 3rd parties as that's another matter entirely -especially with the Nintendo-catering Capcom, Konami, Square, etc)

    He was specifically talking about "16-bit Players" with that comment. Also, nobody knows in real time exactly how much more something will grow, and the market was presently in decline.
    I was purely talking about 16-bit consoles as well: the market was certainly pas the peak for the generation but was surely not saturated yet and had not even reached 80% of the final market totals. (liken it to the NES in 1990)
    But, of course, as time went on that would push more and more into the budget market. (which would have used console/game sales competing as well)

    I have all of them in the footnotes on my page. The quotes Kent uses were quoted in their entirety the first time I brought them up, but the section is at the end of "The Next Generation: Part One" chapter.
    OK, I found it.
    I rather disagree about Nintendo "milking old hardware" given the SNES was only 4 years old by that point, and as history has shown: all popular mass-market game systems tend to last close to 8-10 years before they fall from the budget market and at least 6 years before the fall from major mass-market status unless other factors intervene. (company goes bankrupt, jumps into the next gen and drops support early, or the market crashes -even then the 2600 was strong in the budget market up to about 1989, so well over 10 years, while the Genesis remained in the budget market through 1999 at least -albeit far less than had Sega supported it)

    But you could argue that Sega was also milking old hardware with add-ons as much as Nintendo was with games, and most would argue more so.

    In hindsight we can see that the mass western console market was going to stay 16-bit centric until 1996 at the very least, slump or none, so riding out that slump and not rushing to the next-gen made some sense. And of course Nintendo had their enhancement chips to push a little while Sega had potential for that as well as the Sega CD to do most of that and more even before the 32x came out. (albeit the CD could have been marketed better, but even by '94 there was a lot of potential to push it further and in the context of hype for new hardware: a long overdue low-cost Duo would have been very significant)


    The account is that SoA was developing Saturn software during the 32X development process. It does not say how much software or how much documentation they had received on the Saturn. The 32X documentation was being sent from SoJ to SoA on the fly though. It would seem reasonable that the Saturn's docs were being relayed in similar fashion as they were being written.
    Those documents would need to be compiled and properly translated of course, and that task very well may have fallen on SoA: it might have been up to them to produce the western dev kits too. (in which case that's one more thing SoA was responsible for: not investing enough in comprehensive translations for good low-level development tools and getting a decent programming library together based on that)

    It would be interesting to know what games SoA had started on though as there weren't very many STI/SoA developed/published games early on. (but a fair amount of 32x stuff)

    Since Bayless left the company at the end of 1994, the evidence suggests that the Saturn's Japanese launch damaged the culture at Sega of America.
    DO you have references stating that SoJ ever planned to NOT push the Saturn in 1994? From everything I've seen Sega wanted it out ASAP and had even been pushing it significantly earlier than the actual November release date.
    It seems like Nakayma was skeptical that the Saturn would meet the planned deadline, but that's no indication that it was PLANNED to be released any later than fall of 1994. (especially as they wanted to beat Sony to the market -which they barely did)

    So either that was mis-communication on Nakayma's part, or things changed drastically for SoJ at the time.
    That would be really interesting to know.

    Using sales to retailers was an "innovation" of Sony, not Sega. Sega always focused on sell through to consumers. This would be an oddity if the 600k figure in the US that McFerran claims but does not cite was sales to retailers. I also never saw stacks of 32X boxes anywhere, and Kent's book goes into some detail at the lukewarm reception by retailers in general.
    You mean citing sales in terms of units shipped, not units sold at the consumer level?
    I'd gotten the impression that many, many companies did that including Sega for press-released sales figures, especially if they had to release figures without full returns on sales charts for all distributors. (granted, knowing the real values internally would be critical and that was one major problem with Atari in the early 80s) And Pettus explicitly claims that SoJ had been using such figures to represent the Saturn in Japan.
    Pettus also claimed the intial shipment for fall of 1994 was only 200k, not 600k, and pre-orders supposedly exceeding that 200k. (while 600-700k has been noted elsewhere as an approximate total production/sales run for the 32x)

    You may never have seen stacks or bargain bins full of 32Xs, but many others have accounts to that effect from 1996.

    Not a bridge to the Saturn, a bridge to the 32-bit generation. The cost effectiveness of the add-on was roughly similar to buying a Super Nintendo and would have consistently brought "32-bit" gameplay if given more time.
    A bridge to Sega's 32-bit generation and a bridge to the Saturn would be the same thing: anything else would conflict with the Saturn and mean competing products reducing adoption rate of the newer consoles.

    As for how well they would have handled adaptations of Saturn or PS1 games, that's something we will likely never know. Though Spear 3D running on the 32X like it does, and the other tech demos we have seen, shows a lot of promise for the time.
    I hardly see how Spear3D is a good benchmark for the 32x.. MAYBE Yetti 3D, but even that is a bit less promising. (if nothing else pointing to ray-casting engines being preferable by far)

    But for actual PSX/PC/Jaguar/3DO/Saturn games, yes all of them from the mid 90s would be very tough to port to the 32x for the sheer reason of RAM regardless of processing power. Even the PSX/Saturn had a hard time handling many PC games because of that and in many cases had to cut out a lot or have the game engine totally remade from the ground up as with Duke Nukem and Quake for the Saturn iirc.
    So 32x exclusives and ground-up customized engines would be the only real options for the 32x, and porting polygon/texture heavy games to the 32x would also not be ideal. (there's many games on the PSX and Saturn that could have been much more competitive on the 32x, DOS PC, or Jaguar using ray casting/height mapping to off-set the use of polygons in many areas -ie using mixed renderers with sparing use of polygons and cutting back on textures -especially due to ROM usage)


    I think Bayless is coming from the perspective of what he knew the 32X was designed to do and comparing that to the launch games. In that respect he's right, his phraseology just pisses all over the quality of the library though.
    In some respects (like polygon count) things likely weren't going to get hugely better than what the 32x showed, especially if things appearing on other 32-bit consoles started getting pushed.
    The polygonal portions of Zyrinx's demo are generally unrealistic for a real game: even assuming the 68k had near full resource available, doing all game logic though that would have been limiting still. (albeit at least you had that option unlike the Jaguar where the 68k shares the same system bus with everything else)
    If game logic/AI was kept relatively simple (perhaps a rail shooter), that might have been fairly close to real-life possibilities, but sound would be limited too. (and without PWM DMA, probably not much beyond genesis sound at all -DMA opens things up a lot though)

    I don't think this is an either or thing, I think it's a revenue thing. Sega was losing the Arcade buffer, and that was what caused all of these relatively spontaneous shifts in the way the company had been running.
    They really lost all their buffers by what they did in 1995 though. The 32x was a mixed bag, but the actions taken with the Saturn in 1995 with the early launch to the dropping of support to focus on Saturn at the end of the year were all terrible.
    Sega had a diverse market with the SMS still in a low-budget niche in Europe and strong in Brazil with the mainstream Genesis/MD still big with potential for at least 4 more years phasing into the budget market, the CD might have gone a good bit longer too (the 32x sort of complicated that though), and the Game Gear had lots of potential.
    I'm not sure what they were thinking going all-in with the Saturn as dropping those other platforms certainly didn't help generate revenue, so if anything it made it harder to push the Saturn and of course left them without any real buffer if the Saturn failed. (they screwed up by pulling back on the Saturn too soon as well, albeit somewhat of a different context but similarly harmful for the DreamCast)

    Take a screenshot from an emulator and take a color count, it has been a while, but I don't think Virtua Racing is hitting 256 colors. Metal Head has framerate issues when the textures get complex, Doom was cut down, Space Harrier is great and to heck with anybody who says otherwise, and Cosmic Carnage is a misunderstood stroke of genius that wasn't completed.
    Space Harrier was an old game that should have had a much bigger impact years earlier for the Sega CD... Not a bad addition for sure. (I prefer it to After Burner)
    And if Virtua Racing isn't highcolor then no game on the 32x is.

    Doom was cut-down by necessity and doing better would have been a matter of a lot of hacking and modification, but still very limited without more ROM. (and even then fundamentally limited by the slow ROM bandwidth or trade-offs of on-the-fly decompression -GBA Doom has the advantage of much faster ROM as well as a full 8 MB of ROM used)
    Of course, they could have dumped a port altogether and gone with a new game engine entirely like the SNES version of Doom did.
    Short of a total remake, being able to offload the game logic to the 68k (fitting it within the 64k work RAM) and leaving the 32x to pure rendering would help, unless they managed to do that already. (there's also the audio to consider, and use of DMA would have offloaded a lot of resource going to that too and/or allowed a sample based synthesizer with compressed samples)
    In Chilly Willy's case, he'd be using the Sega CD exclusively: so the ROM issue is gone, but managing to cram the game into little more than 512 kB (you could put some textures in word RAM) is the main challenge. (you also have the 12.5 MHz 68k to work with, plus the 64k wave RAM and PCM chip... you might even offload some stuff to the ASIC rendering to the genesis VDP if you really went for that -namely 2D sprites/objects- optimized for the Genesis palette, of course)

    I don't own any of the EA ports, and only own one of the Akklaim games I think, but I find the rest of the library more compelling than most of the other games that came out at the same time.
    More compelling than the best contemporary Genesis, CD, SNES, 3DO, and Jaguar games? (let alone early Saturn and PSX contemporaries)

    I agree, the time to cancel the 32X was before it was announced not after. Still, I don't think Sega as an organization had a better solution for 1994's revenue problems. It makes me sick that the Genesis and Sega CD software libraries alone didn't create an upswing.
    That all could have been up to marketing though, especially with the CD finally falling to reasonable mass-market prices (a low-cost duo could have helped a lot though), but then there's potential for more SVP games (or SVP add-on) and CD games. (let alone SVP-CD games)
    Who knows how that might have performed, especially Star Wars Arcade on the CD or SVP... Virtua Racing is OK, but the SVP could have been pushed for much more popular stuff, and it's a shame VR wasn't lock-on in the first place, especially if designed to work with the CD as well. (in the latter case you'd have even more options as the SVP could be dedicated to plotting points for the Sega CD would then use to rasterize into polygons -potentially texture mapped- and be faster than either alone)

    You could have all sorts of SVP+CD stuff (including improved compression for FMV), and with Virtua Racing or such an SVP add-on being generally more accessible than the 32x, that could have been a lot more practical. (a bare SVP add-on probably could have been $50-75 at the time depending on how Sega managed profit margins and like SVP games, would have fallen as more were produced)
    Hell, it could have been integrated into the Duo too fo rout of the box Genesis+CD+SVP compatibility.

    And then there's the 2D games on the 32x which could have been done reasonably well on the CD (if not better in some respects), with some exceptions.

    The main limits would of course be color, but that varied a lot in how dramatic it was via composite or RF. (especially given it was all 320 wide which blurs a lot more than 256 wide) Though things were a bit different for Europe with composite being a bit sharper and access to RGB in many cases.
    6 days older than SEGA Genesis
    -------------
    Quote Originally Posted by evilevoix View Post
    Dude it’s the bios that marries the 16 bit and the 8 bit that makes it 24 bit. If SNK released their double speed bios revision SNK would have had the world’s first 48 bit machine, IDK how you keep ignoring this.
    Quote Originally Posted by evilevoix View Post
    the PCE, that system has no extra silicone for music, how many resources are used to make music and it has less sprites than the MD on screen at once but a larger sprite area?

  8. #8
    I remain nonsequitur Shining Hero sheath's Avatar
    Join Date
    Jul 2010
    Location
    Texas
    Age
    42
    Posts
    13,313
    Rep Power
    130

    Default

    Quote Originally Posted by kool kitty89 View Post
    The best example of a multiplatform 2D game making good use of the 32x would probably be Blackthorne, but that doesn't really show off that much other than color and animation (no scrolling, not much going on on-screen etc). ...
    I was disappointed with it at first, the extra horizontal resolution from the SNES game was just used for the menu. I was shocked to see how many colors were actually being output (>256 32X vs 85 SNES). I would have thought the two versions were relatively close even eyeballing from emulation. I thought the sound was okay in both versions, though I haven't played the SNES version on real hardware yet.

    Quote Originally Posted by kool kitty89 View Post
    I really don't see it though, but in any case: such a short time period pretty much killed any chance for reasonable hardware support, so going simple with hefty off-the-shelf CPU grunt was a good option in that respect....
    It's just his opinion. History supports the idea that specialized chips with high level optimization were used more. This is the same argument as quads versus triangles in my mind. It is a chicken and egg argument. Did one idea take over because it was intrinsically better, or because more companies used it? The answer is a matter of opinion, though one can point at software performance in published games and draw some objective conclusions.

    Quote Originally Posted by kool kitty89 View Post
    Jupiter was a mass market name applied after the fact: Sato's mars design (AKA Super Mega Drive) was totally separate from the proposed direct Saturn derivative from everything I've seen. In fact, I've never seen any very useful info on Sato's mars design beyond Pettus's simplified quotes from Joe Miller regarding "128 colors."
    If the "Jupiter" (ie totally incompatible with MD but directly forward compatible with Saturn) design was ever on the table, it was a big mistake to drop it, but given Joe miller's comments about Sato's Mars being rather weak and barely more than a "Genesis with double the colors" seems to totally contradict that.
    I really don't see how the 32X and the Saturn aren't compatible from a development perspective. Surely an assembly coded 32X game could be relatively easily ported to a Saturn CD-ROM and even made better in the process. I for one would not have been interested in buying a "cheaper" console only to have to buy a completely new one (Saturn) to play the full fledged games. The 32X enhanced Genesis and Sega CD games *and* gave "32-bit" gameplay. The Neptune, as theorized, would have provided cart based Saturn games which would have been immediately obsolete on a CD-ROM based system. I don't get the infatuation.

    Quote Originally Posted by kool kitty89 View Post
    I don't see where in the article it implies that SoA thought the Saturn wasn't goign to be out anywhere by 1994. It does say that Nakayma was possibly skeptical about the 1994 deadline being met, but that's hardly definitive. (unless I overlooked something)
    I might not have quoted in enough detail. Bayless goes into some detail about a consensus at SoA that the 32X was the right direction. He then states that the Saturn's Japanese launch ruined that perception.

    Quote Originally Posted by kool kitty89 View Post
    No, I meant that the Saturn was still in development during early 1994 in the context of Nakayma doubting it wouldn't be ready for a release that year.
    Having a design still yet to be completed can still have the hardware pretty much set but require a LOT of final work to eliminate bugs and actually make it production quality.
    That much should be obvious. If the Saturn wasn't still being developed they could have started manufacturing much earlier.

    Quote Originally Posted by kool kitty89 View Post
    He has no grounds for criticizing the Saturn in any case though as it was universally more powerful and flexible than the 32x and no more difficult to program for in apples to apples comparisons. (in fact, generally EASIER due to the greater system resources)
    From a cost standpoint there's a big difference, but he wasn't arguing that.
    I'm not justifying his statements, I'm qualifying them. Regardless of whether Bayless is right, it is important to explore his perspective due the the relative lack of primary sources on this topic.

    Quote Originally Posted by kool kitty89 View Post
    So does the 32x and every other platform ever made that renders 3D as such. That's how polygonal 3D works, 2D primitives warped to represent 3D space using based on plotted 3D coordinates.
    Try posting that on any mainstream media website regarding the system and prepare to get shouted out.

    Quote Originally Posted by kool kitty89 View Post
    The ONLY reason the Saturn and 3DO get criticized for "not being 3D enough" if the use of quadrilateral primitives going against common convention, but Sega's own Model 3 did the same thing as is praised as the greatest 3D arcade system of its time.
    I find the criticism of quads absurd. I can't see it any other way than an apples to oranges comparison. Perhaps somebody can argue mathematically that triangles were somehow more efficient than rectangles.

    Even then it sounds no different to me than coders claiming that the SNES CPU was more efficient than the 68000 because of how many lines of code it took to do similar operations. I suspect one's background and experience, and the environment the original game was created for, factored much more heavily than a simple comparison.

    Quote Originally Posted by kool kitty89 View Post
    I'm not sure what memory timing issues that would be, but the 32x had a ton of bottlenecks to deal with in that regard...
    I'm also not sure what "like the 32x" would mean either: if you meant software based, then no, the Saturn would have been much weaker then and no games would have looked any better. ...
    Unless you're saying that instead of ANY significant hardware acceleration, the Saturn should have been built around a CPU+framebuffer set-up with simple sound as well...
    He just seemed to rattle off a series of popular complaints with no personal knowledge of the Saturn to me. That is the kind of speculation I don't have the liberty of delving into in an objective review of a very subjective article though.

    Quote Originally Posted by kool kitty89 View Post
    Well, then that's a patently false and misleading statement as 320x224x256 color (from 15-bit RGB) was NOT a major thing by that point and was pretty much what mid-range PC games had been pushing for several years at that point. (and well below that of the Jaguar and 3DO) It's also not far from what the SNES had been pushing either, albeit that was using 16-color tiles and sprites with 256 indexes from 15-bit RGB and at 256x224/239.
    I didn't think his "colors per pixel" comment made any sense at all, and he definitely did not elaborate.

    Quote Originally Posted by kool kitty89 View Post
    You mean by the end of 1993, not '94 right?

    And no, while I'm sure pure software rendering using CPU resource was NEVER a serious consideration for the Saturn (and really shouldn't be for any non-rushed system) software rendering is ALWAYS an option: the Saturn could and DID do it better than the 32x, but it had a lot of added hardware to help out and that's where it shined.
    Hmm, yes I did mean the end of 1993 there. Bayless account makes any possibility of the Saturn not having a solid hardware design by the beginning of 1994 improbable. But the Saturn's actual design and launch that year in Japan already proves that.

    Software rendering is an option if the hardware manufacturer provides the info or the developer strains that relationship by reverse engineering. Obviously most developers are going to go ahead and use all resources available if that is an option.

    Quote Originally Posted by kool kitty89 View Post
    If the SH2s were added in a redesign, that likely was the FIRST change as that would be the only off-the-shelf addition and if the redesign began in late fall of 1993 (months before Mars was brought to SoA), the SH2s easily could have been part of the Design by then, even if not fully prototyped yet and even if the other hardware was also being modified.
    I would have to see significant primary evidence that the Saturn had the SH-2s added after its initial announcement in mid 1993. I can see a later stage DSP addition/modification, or even the VDP's and RAM being rethought, but not the SH-2s. The existence of the SH-1 in the Saturn only demonstrates Sega's willingness to stick to code compatible central processors in my opinion. This is similar to why the Sega CD uses a 68000 instead of using something arguably just as cost effective but incompatible with the Genesis CPU.

    Quote Originally Posted by kool kitty89 View Post
    The only software rendered portion might have been 3D if VDP1 wasn't there at all, but VDP2 already is a hefty chunk of hardware acceleration ... but it's pretty likely that there was something in place of VDP1... VDP2 was the real monster in the Saturn though, and if you took that out and instead integrated simple video generation ... into VDP1 and dropped VDP2's RAM it would have been significantly cheaper but much worse at 2D.
    I do think that SoJ made absolutely certain that the Saturn would have no deficiencies at 2D. No matter how arguable I find the Saturn's 3D prowess, I consider its 2D dominance that generation as close to fact as these things come.

    Quote Originally Posted by kool kitty89 View Post
    That's false, the 2nd SH2 is only a small part of it: it is indeed tough to really get good performance out of the slave CPU and generally needs to be used for some specific tasks to make efficient use of the shared bus. (at best you probably get 60% performance if that) But it's added resource, they don't force you to use it.
    This always seems to be comparing the dual SH-2s to a computer system or workstation that would use dual CPUs to simultaneously work on separate code. Everything I've seen says that the Saturn was not designed for that purpose, instead the second SH-2 was included to work while the first was busy.

    I consider this the same idea, though of course not the exact same implementation, to how modern PCs use dual core processors even when software is not optimized for them. Even though the real time rendering performance would improve in audio/video/3D editors if the software was dual core optimized, there is still a significant performance increase (nearly double in most cases) from a single core setup running at the same speed.

    Quote Originally Posted by kool kitty89 View Post
    The main problems are not related to the CPUs at all though, but the general hardware and software development tools not being aimed at high-level programming, and due to the hardware, even with later high-level tools it was far less efficient than the PSX which was designed specifically for that like most later consoles (Dreamcast is WAY up there but GC, Xbox, and 360 are too).
    I agree, except that I think Sony basically created this problem by marketing high level development kits to third parties in the first place. Revolutionaries at Sony goes to great lengths to basically make it look as though Sony was "smart" and everybody else was holding the industry back for just that reason.

    Quote Originally Posted by kool kitty89 View Post
    And then there's the use of quad rasterization on top of all that.
    But the generally poor early documents (even for low-level programming they were poor and omitted documentation of key features like the geometry DSP) and improved tools were late to the game. (and even if they had decent comprehensive high and low-level tools from the start it wouldn't have been nearly as easy to get good 3D performance out of as the PSX, but at least a lot better than it was)
    The early software does_not demonstrate this rumor. Exclusive PS1 games seem to demonstrate the system's relative 3D prowess, but even they have exclusive Saturn counterparts (High Velocity, VFR/2, Cyber Speedway, WingArms) that gray that area completely.

    It is possible that Sega, as Pettus claims, held back information from third parties because that was tradition. I have yet to see any primary source that demonstrates such a mentality though.

    Quote Originally Posted by kool kitty89 View Post
    If it had no VDP1, SH2s, or geometry DSP, then yes pretty much limited to model 1 ports and the like ... all of which would likely be perhaps moderately close to what the 32x was outputting. ...
    This alone discredits the idea that Sega was seriously considering such a solution after 1992. Something capable of porting Daytona and Virtua Cop (at a minimum) would have been on the table as soon as their arcade counterparts started development.

    Quote Originally Posted by kool kitty89 View Post
    But it was very likely there was some VDP1 for a blitter/sprite engine as I said, albeit in that context it likely had no 3D modes ... and the DSP likely was also missing ... you had a good bit more help with the addition of hardware scaled/rotated sprite/object rendering and also affine line rendering. ... Add that DSP and you've got a lot more to work with ...
    I keep seeing the DSP referred to as some sort of helper chip for 3D, but the docs point to it being inserted as a special code processor, especially for DMA. I think there is some disparity as to whether the DSP was poorly documented versus Sega providing every example of special processes the chip could handle. I also don't trust the idea that the DSP was primarily for 3D calculations, as opposed other smaller processes involved in total system management. One of Sega's 1995 documents that I have seen claims that is exactly what the DSP was for.

    Quote Originally Posted by kool kitty89 View Post
    I'm not sure what the hell you're talking about? The Sega CD's added CPU and RAM have nothing to do with load times, but the 16 kB CD-ROM cache does (compared to the cacheless PCE CD), that would have been true if a cheap simple embedded MCU was used in place of the 68k.
    Could the Genesis alone have loaded data off the CD and processed it without *any* hit to load times?

    Quote Originally Posted by kool kitty89 View Post
    Likewise the Saturn had a hefty 512 kB CD-ROM cache/buffer that the SH1 worked in (but again, like the CD's CPU it was way overkill) vs the 32 kB cache of the 3DO and PSX. (that's the same reason the CD-Z loads faster than the Neo Geo CD, and I think the Wondermega/X'Eye also have enhanced caching over the standard MCD)
    I'm not sure whether the X'Eye has more cache, I recall its load times not being significantly faster anyway. The X'Eye was supposed to have a much more optimized CD-ROM drive to eliminate "data access time." Mine certainly never impressed me over my Sega CD Model 1, I haven't spent much time with my SCD2.

    Quote Originally Posted by kool kitty89 View Post
    There's also other things like on-the-fly data decompression to speed up loading, but that's possible on all those systems (and the PSX has a hardware decompression engine for that very purpose), and the one final way to speed up loading is putting tons of redundant data on the disc to greatly reduce seek times and allow linear loading. (but that only works if you have enough extra space on a disc or are willing to use more discs -that's also the main reason pirated Dreamcast ISOs require lots of seeking compared to GD-ROMs)
    NEC interviews around the time of the Sega CD's development were almost exclusively promoting newer compression solutions for the PCE/TG-CD. My only experience with the PCE-CD experiencing frequent or long load times is with Altered Beast CD. It was a port of the 4mbit PCE HuCard with enhanced audio and some other (significant in my opinion) graphical enhancements. The game loaded though, about three to four times per level, totally freezing the action. NEC Avenue, for whatever reason, couldn't eliminate these loads in early PCE-CD games. Apparently they simultaneously worked on more advanced compression solutions and the System Card updates.

    Quote Originally Posted by kool kitty89 View Post
    But the Jaguar failed to really get mass market attention and general support...
    As it was, the 3DO outsold it by more than an order of magnitude and it really failed to reach critical mass on the mainstream marketplace. ...
    These were by far not foregone conclusions at the time.

    Quote Originally Posted by kool kitty89 View Post
    I'm still not sure what that means as both Sega and especially Nintendo had older consoles still on the market ...
    Welcome to the wonderful world of investment marketing. Since all of these are US sources, and we're speaking of 1994, I would suppose every comment is US centric. At the least they only make special mention of the world markets (still plural at the time) in parenthetical statements.

    Quote Originally Posted by kool kitty89 View Post
    Sega had hits more spread out than Nintendo, but a LOT of software sales and I've not seen any comprehensive figures that pointed to Nintendo being strong in that area, especially if you exclude Japan entirely.
    Nintendo and their major third party supporters dominated the "million seller" lists for that generation.

    Quote Originally Posted by kool kitty89 View Post
    If the software wasn't selling in spite of the quality there's 2 main reasons: marketing wasn't right and/or the game wasn't catering well enough to the mass market. (difficulty, gameplay mechanics, art style, etc) And then comes brand recognition, but by the mid 90s Sega had that down in North America and in most of Europe as well, albeit not the established series of Nintendo. (Mario, Metroid, Zelda, etc vs Sonic, Phantasy Star, and a few others to a lesser extent -not getting into 3rd parties as that's another matter entirely -especially with the Nintendo-catering Capcom, Konami, Square, etc)
    I would put brand recognition as the prime candidate for why Sega and their major third party supporters didn't see higher sales on their original IP.

    Quote Originally Posted by kool kitty89 View Post
    I was purely talking about 16-bit consoles as well: the market was certainly pas the peak for the generation but was surely not saturated yet and had not even reached 80% of the final market totals. (liken it to the NES in 1990)
    But, of course, as time went on that would push more and more into the budget market. (which would have used console/game sales competing as well)
    I suspect that if we could compare worldwide Genesis/MD sales to SNES/SFami sales going into 1994 the situation would look dire for Sega. We've already seen their financial situation, which alone mandated radical action.

    Quote Originally Posted by kool kitty89 View Post
    But you could argue that Sega was also milking old hardware with add-ons as much as Nintendo was with games, and most would argue more so.
    Cash cows are just that. That is why Sega focused on new hardware and software and Nintendo focused on software.

    Quote Originally Posted by kool kitty89 View Post
    Those documents would need to be compiled and properly translated of course, and that task very well may have fallen on SoA: it might have been up to them to produce the western dev kits too. (in which case that's one more thing SoA was responsible for: not investing enough in comprehensive translations for good low-level development tools and getting a decent programming library together based on that)
    Yes, that is one thing that I left some detail out of in the review. Bayless describes "technical Japanese" as being an english/japanese hybrid that they had to outsource for translation. He also describes how hilarious the translations would always be.

    Quote Originally Posted by kool kitty89 View Post
    DO you have references stating that SoJ ever planned to NOT push the Saturn in 1994? From everything I've seen Sega wanted it out ASAP and had even been pushing it significantly earlier than the actual November release date.
    It seems like Nakayma was skeptical that the Saturn would meet the planned deadline, but that's no indication that it was PLANNED to be released any later than fall of 1994. (especially as they wanted to beat Sony to the market -which they barely did)
    Bayless paints this as an assumption on SoA's part. The entire company in the US thought the 32X was going to be the only thing on the market until 1995. McFerran also states that SoJ had a "consensus" that Saturn should be emphasized above all else.

    Quote Originally Posted by kool kitty89 View Post
    So either that was mis-communication on Nakayma's part, or things changed drastically for SoJ at the time.
    That would be really interesting to know.
    According to this interview it was a lack of communication on Nakayama's part exclusively. As a result of his actions the whole of SoA got excited about the 32X while SoJ provided exhaustive support and focused on getting the Saturn out by the end of 1994.

    Quote Originally Posted by kool kitty89 View Post
    You mean citing sales in terms of units shipped, not units sold at the consumer level?
    Everything I have seen shows Sega pointing to units sold through to consumers. This was even a major point of contention when the Dreamcast launched. When Sega reported one million units sold in less than two months people familiar with Sony dismissed it as production numbers not sales. But even EA knew they had sold that many units to consumers by November, and shifted their PR stance as to why they weren't going to support it.

    Quote Originally Posted by kool kitty89 View Post
    You may never have seen stacks or bargain bins full of 32Xs, but many others have accounts to that effect from 1996.
    This point is regarding 1994's US launch exclusively. Apparently McFerran accepts that 600k 32X sold at and after launch in the US alone, and something similar happened in Europe.

    Quote Originally Posted by kool kitty89 View Post
    A bridge to Sega's 32-bit generation and a bridge to the Saturn would be the same thing: anything else would conflict with the Saturn and mean competing products reducing adoption rate of the newer consoles.
    Not necessarily, especially not if marketed correctly. The Sega CD was already "bridging the gap" between high end PC games and 3DO games. It allowed Genesis owners a relatively low cost solution to play games they otherwise could not without significant investment. Somebody privileged with a mid range PC at the time would have a hard time seeing that.

    With that said, I do not know of a single example of a company releasing forward compatible hardware. I especially cannot figure how a cart based "Saturn" with no backward compatibility would have benefited anything.

    Quote Originally Posted by kool kitty89 View Post
    I hardly see how Spear3D is a good benchmark for the 32x.. MAYBE Yetti 3D, but even that is a bit less promising. (if nothing else pointing to ray-casting engines being preferable by far)
    I was thinking of Yetti, though Spear 3D would have been a welcome launch addition to the 32X.

    Quote Originally Posted by kool kitty89 View Post
    But for actual PSX/PC/Jaguar/3DO/Saturn games, yes all of them from the mid 90s would be very tough to port to the 32x for the sheer reason of RAM regardless of processing power.
    Sales heals all wounds. If there's a will there's an engine.

    Quote Originally Posted by kool kitty89 View Post
    In some respects (like polygon count) things likely weren't going to get hugely better than what the 32x showed, especially if things appearing on other 32-bit consoles started getting pushed.
    The polygonal portions of Zyrinx's demo are generally unrealistic for a real game: even assuming the 68k had near full resource available, doing all game logic though that would have been limiting still. (albeit at least you had that option unlike the Jaguar where the 68k shares the same system bus with everything else)
    So I take it you think the Scavenger team demo lied that all of those demos were running 100% real time on a 32X? Do you have any sources for this? I have found Scavenger to be one of the more reputable developers from the time, and very hard core into creating extremely unique engines.

    Quote Originally Posted by kool kitty89 View Post
    More compelling than the best contemporary Genesis, CD, SNES, 3DO, and Jaguar games? (let alone early Saturn and PSX contemporaries)
    For me? Definitely, I was an arcade game whore. After Burner was my "Super Mario Bros", I wouldn't be a gamer without it. Shadow Squadron and Virtua Racing were my passionate obsessions in 1995 until I started playing Saturn games that Fall. I played Genesis and Sega CD games to be sure, and kept track of SNES games at friends houses (and played 3DO and Jaguar in kiosks) but I was more than entertained by 32X software.

  9. #9
    Hero of Algol kool kitty89's Avatar
    Join Date
    Mar 2009
    Location
    San Jose, CA
    Age
    30
    Posts
    9,724
    Rep Power
    63

    Default

    Quote Originally Posted by sheath View Post
    I was disappointed with it at first, the extra horizontal resolution from the SNES game was just used for the menu. I was shocked to see how many colors were actually being output (>256 32X vs 85 SNES). I would have thought the two versions were relatively close even eyeballing from emulation. I thought the sound was okay in both versions, though I haven't played the SNES version on real hardware yet.
    Interesting: is that over 256 colors counting the score bar, or just the gameplay screen? I's imagine the full screen as otherwise it would imply the game was rendered in highcolor which would make little sense given that it's not clipped to 204 lines or less.
    The status bar is the only thing in-game rendered by the Genesis.

    What else is odd is that the VGA DOS game doesn't look nearly as good when it should look at least as good as the 32x game... maybe they stuck to the default VGA palette for some reason (it's not a restriction, but many, many developers used it for some odd reason).
    In any case, the VGA game hardly looks better than the SNES game.

    The SNES version obviously has to work around the 256x224 resolution as well as the 15 colors per tile/sprite limitation plus the 8 palettes for sprites and BG. (ie 8x15 color palettes for the sprite and another set for BG all selected from 15-bit RGB -same as the 32x) So a lot better than the Genesis, but limited compared to a 256 color bitmap display like VGA or the 32x. The SNES could have used mode 3 with 256 colors per tile, but that would use up a lot more memory and also hinder animation due to increased VRAM DMA time needed for updating.
    However all of that is a bit moot given the 32x used 3 times the ROM as the SNES (albeit a year later) with 3 full MB to the SNES's 1, so that's a massive difference right there for both audio and video.

    As for Sound, the 32x version has pretty noisy PWM and fairly low quality samples I think: I wouldn't say it sounds better than the SNES overall, but it's debatable to a degree. The 32x's sound system is more limited than the SNES's for sure, but in particular it's very limited by any officially released game due to the lack of working DMA in the dev kits, so ALL the PCM/PWM playback had to be handled in software using a ton of resources. (otherwise the only resource needed would be for software mixing multiple channels -again DMA and no DMA is the difference from the slave SH2 managing 32-64 channels vs 4-8 channels at 22 kHz)
    OTOH that's even true of the PSX, Saturn, and Dreamcast, especially in terms of the audio hardware: simple DMA based audio hardware perhaps with hardware ADPCM decoding or simply using a fast CPU/MCU for that instead. (in all cases the DSP could pretty much be dropped... though in the DC's case the ARM7 would have really come in handy for MP3 decoding to open up a ton of GD-ROM space normally used for CD-DA without having to use CPU grunt to do so -and much better/higher compression than ADPCM)

    It's just his opinion. History supports the idea that specialized chips with high level optimization were used more. This is the same argument as quads versus triangles in my mind. It is a chicken and egg argument. Did one idea take over because it was intrinsically better, or because more companies used it? The answer is a matter of opinion, though one can point at software performance in published games and draw some objective conclusions.
    Not used more, but fundamentally better and FAR more cost effective. There is a middle ground and that would be a powerful and flexible coprocessor that's still more focused than a CPU but a lot more flexible than a dedicated blitter or GPU (like a DSP or the RISC "GPU" in the Jaguar), but really it's probably most efficient to do a bit of both.
    In the case of the PSX it was fully dedicated hardware with the main flexibility being the CPU while the GTE was solely for geometry calculations and the GPU was only really useful as a blitter/rasterizer: had Sony made the GTE a bit more flexible and not fully fixed-purpose for vextex plotting, it could have been very useful for other operations. (perhaps ray-casting/voxels/height maps even in the manner the SVP or Super FX were, but much faster -maybe closer to the Jaguar's GPU) The vertex performance would have dropped most likely unless they added more in general (and added cost), but that might not have been so bad, especially given vextex generation usually wasn't the limiting factor of the PSX (texture mapping or rasterization speed were iirc, maybe game logic timeouts in some cases). Interestingly enough, the GTE in the PSX was one area that PC graphics cards took a long time to push in hardware as most early 3D cards (even the high-end ones) required the CPU to handle the 3D math while the accelerator handled the rasterization/drawing/textures/shading/filtering/perspective correction/etc. (I think hardware Z buffering came before hardware vertex plotting -which is actually something the jaguar focused on as well, albeit without fast texture mapping or hardware rasterization)

    As to polygons: trips vs quads is a whole other issue too: in the real world triangles just make a lot more sense and they did 20 years ago or 30 years ago too, or longer. The algorithms and general math used and developed by humans is tied very heavily to triangles such that pushing quads really complicates that among other things.
    Then there's the general flexibility of triangles vs limitations of using quads: you can easily pair triangles for quadrilateral surfaces (let alone larger polygons), but you can't warp a quad into a polygon (ie overlap 2 points) unless it's solid shaded: doing so with a gouraud shaded or texture mapped quad will horribly warp the texture or screw-up the shading.

    There's a reason pretty much all (polygonal) 3D games prior to hardware acceleration used triangle rendering (including the 32x and jaguar) and also a reason 3D workstations and early 3D arcade boards predominantly did as well. (I'm not positive, but I'm almost sure that the Hard Drivin' boards, Model 1, Model 2, and Namco System 21 and 22 all used triangles exclusively and the ST-V or maybe some arcade applications of the 3DO were the first arcade systems to use quad based rendering)
    The early PC 3D accelerators also pushed polygons with the solitary exception of NVidia's NV-1 and canceled NV-2 chipsets. (the 1995 ATI and S3 chipsets -namely Rage and the low-end ViRGE- were triangle based, and of course so was the Voodoo)

    Sony wasn't setting a trend, but solidifying it: the bigger thing that the 3DO, Saturn, PSX, etc were pushing was polygonal 3D over the main competition on software rendering consoles (namely PCs) ray-casting based height mapping (including Wolf 3D, Doom, Duke 3D, and voxel engines). There was every possibility of hardware acceleration aiding ray-casting as well, but it didn't happen that way.

    As it was, it's almost certain that a major reason the Saturn and 3DO used quads was purely due to engineering time/cost due to the rectangle based "sprite" rasterizer/blitter (like what the Sega CD has) being just a step below a 4-point warped quad rasterizer while a triangle rasterizer would take a fair bit more work. (of course they could have droped the warped quad rasteriztion logic altogether and added a fast enough CPU to handle the added overhead instead, or a suitable DSP -line by line rasterization like all software polygon renderers but with texture mapping and shading in hardware rather than plain solid color line fill)


    I really don't see how the 32X and the Saturn aren't compatible from a development perspective. Surely an assembly coded 32X game could be relatively easily ported to a Saturn CD-ROM and even made better in the process. I for one would not have been interested in buying a "cheaper" console only to have to buy a completely new one (Saturn) to play the full fledged games. The 32X enhanced Genesis and Sega CD games *and* gave "32-bit" gameplay. The Neptune, as theorized, would have provided cart based Saturn games which would have been immediately obsolete on a CD-ROM based system. I don't get the infatuation.
    Yes, but any such games would look crappy compared to what the Saturn was capable of.
    It should have been fairly easy to port some 32x games to Saturn, but applying significant enhancement would be another story: the simplest thing would be speed improvements from the fast main RAM compared to very slow ROM on the 32x and SDRAM 1/2 the bandwidth of the Saturn and no need for on-the fly decompression, plus the ~25% faster CPUs. But even for a straight port you had the Genesis hardware complicating things and the sound engine would need to be completely re-done. (for the Genesis 2D stuff you'd need to replace that with Saturn BGs and sprites) Not to mention games running the game logic/AI on the genesis CPU. (let alone the Sega CD)
    Any polygonal games would be pretty stuck with software rendering... maybe they could have tweaked the games to do line by line shading and texture mapping using VDP1, but not hardware rasterization. (and unlikely that the DSp would be used either, especially due to the poor documentation)
    If it was a ray-cast game like Doom or Metal Head, that would be all software anyway so not too much of an issue. (I'm assuming the DSP isn't flexible like the one in the SVP)
    Then there's the fact that you might end up wasting the slave SH2 in the Saturn if a 32x game dedicated the slave SH2 to handling audio (as was not uncommon, and the PWM sucked a ton of resource due to the lack of DMA support), so the game would need to be modfied to take advantage of the added CPU as well. Then you'd also have highcolor vs 256 color 32x rendering on the Saturn, unless you pulled back to the 256 color mode.

    So some 32x games could have simple/sloppy ports to the Saturn, especially with no or minimal use of the Genesis hardware, but they wouldn't usually be a ton better without further optimization.
    They probably should have done that for the better 32x games for sure and it's pretty odd that they didn't: especially Doom, which was still a killer app in late 1995 and should have benefited significantly even from a simple port -quite possibly running better than the Saturn's later port which I'm almost certain uses polygons like the PSX game and thus is slower than raycasting. (the content should have been similar to the 3DO game at least, but fullscreen, smoother than the 32x, and maybe in high-res mode -if they put a bit of work into it they should have been using both SH2s and using VDP1 to handle shading like the Jaguar did -not as smooth as the jaguar as it's only highcolor rendering)
    Then there's VIrtua Racing (more of a filler, granted), Cosmic Carnage, Metal Head (likely higher framerate, longer draw distance, and higher-res textures if nothing else -plus audio), Star Wars Arcade (definitely), and you had some others, but I'm not sure how many were even out only he 32x by mid 1995. (VF remix predated VF32x obviously) They wouldn't have merited the Saturn being launched in May, but would have been nice additions nonetheless, especially alongside Kolibri, Tempo, Chaotix, etc or for a more moderate early launch like August (3rd party games are more dependent on getting interest in Saturn development early on). Unless Sega could have ramped up Saturn production and collectively planned the may launch with a dozen plus games, a healthy number of systems, developers and retailers notified months ahead of time, western Saturn development starting by fall of 1994, etc. (some of that dependent on SoJ, but otherwise the only remaining drawback would be the $400 price tag, unless they were willing to dump that to $300 or perhaps $350)

    But a major point would be that the Saturn would not have helped developers transition to the Saturn in general: it would have promoted tendencies that would often do nothing but not promote the Saturn's best performance. The specifics and quirks of the architecture in the Saturn, especially making good use of the VDPs. (the main thing the 32x would not do at all)

    Had the Saturn been just a faster a CD based 32x with fast RAM and better sound system (maybe a powerful DSP) that would have been very different and Sega could easily have undercut Sony's price. but that was hardly the case. (it wouldn't have been as capable as the PSX for sure and weaker than the Saturn in polygonal 3D as well in many cases -or at least weaker than VDP2 heavy games- but the cost would have been less... albeit they could have had an even more efficent machine -at least a good texture/polygon pusher- if they put work mainly towards a low-cost optimized GPU/blitter that would work efficiently on a single bus)
    But from the purely CPU based area, waiting a couple months and using a 66 MHz SH3 could have been amazing. (the SDRAM they were using was already rated for at least 66 MHz, but underclocked due to the slow system, so you'd automatically more than double the bandwidth, or more than quadruple it if the SH3 uses 64-bit DMA -not sure if it does, but the SH4 does and most contemporaries, I'm also not sure it has an FPU which would also be very significant)
    More costly than dual SH2s from a component cost perspective, but a much smaller and less cluttered PCB and enough CPU resource to allow a lot of good stuff even with something as simplisitic as the 32x VDP. (to save cost further, you'd probably want such a VDP sharing the main bus and using 64-bit DMA to minimize penalty to the CPU) As it was, again, there was a lot of lesser changes to the Saturn that could have been done to trim unnecessary things and optimize more. (and it would seem a waste to throw all that work in the trash -not developing the Saturn at all would be another thing though )

    I might not have quoted in enough detail. Bayless goes into some detail about a consensus at SoA that the 32X was the right direction. He then states that the Saturn's Japanese launch ruined that perception.
    Yes, but why didn't SoA know the Saturn was going to launch then?
    In that context they would have either canceled the 32x or tried hard to put a positive spin on the Saturn.
    HAD the Saturn been held back, the 32x definitely would have made more sense than it did, maybe not necessary, but more reasonable. (especially if the Saturn was held to allow more modifications, refinement, optimization, and especially put together good development kits)

    That sounds like a prime example of miscomunication due to there being every indication of SoJ wanting the Saturn out ASAP and aiming at a 1994 release, especially to meet (or preferably beat) Sony in the market. (you mention Nakayma was worried about the Saturn being ready for 1994, but did he ever tell SoA that they were or were not planning to release it in 1994?)

    Try posting that on any mainstream media website regarding the system and prepare to get shouted out.
    Mainstream is the definition of ignorant consumer... at least if Xbox 360 sales are any indication.
    Anyone with any technical understanding would not disagree though.
    Hell, in the most fundamental sense: EVERYTHING is 2D on the basic level as it's projected to a flat/2D screen (and stereoptical 3D is just dual 2D). So until we have holographics projections, no, there's no "real" 3D as such.

    But on the most fundamental level: 3D in computers and video games implies the use of points plotted to 3D space (the main reason ray-casting isn't usually considered real 3D), somthign quad and triangle engines do, but somthign a fully "2D" game could also do. (ie you could have a fully scaling/animated or scaling/rotation "sprite" based game that actually maps out the objects in 3D space and then renders the "sprites" accordingly -I'm not sure if Wing Commander does that, but that's sort of the idea)

    I find the criticism of quads absurd. I can't see it any other way than an apples to oranges comparison. Perhaps somebody can argue mathematically that triangles were somehow more efficient than rectangles.
    Yes, yes they have as I briefly described above along with the severe restrictions it puts on models in conventional rendering. (you also need to totally redesign a renderer to switch beteen the 2 in hardware... albeit you COULD have both, but it's never been done in hardware AFIK or even in software for any commercial games)

    http://sega-16.com/forum/showthread.php?t=11678
    I think I've mentioned it before, but I started a thread on it which also included some mixed arguments from Atariage. (including AtariOwl, one of the few developers who actually prefers quads and is actively using a purely quad based renderer in his homebrew jaguar game)
    Note I hadn't realized the model 3 used quads yet.

    Even then it sounds no different to me than coders claiming that the SNES CPU was more efficient than the 68000 because of how many lines of code it took to do similar operations. I suspect one's background and experience, and the environment the original game was created for, factored much more heavily than a simple comparison.
    No, not purely lines of code, but cycle times: the 65815 has very fast instructions compared to the 68k, but also simpler instructions while the 68k's are more powerful, so it depends on the case: for operations that cater well to the simple instructions the 650x arch has a huge clock/clock advantage over the 68k, but for stuff that makes use of the more powerful instructions of the 68k, it can very well have an advantage too. (in most games the former was the case I believe) The same is true for the PCE's CPU, albeit it's considerably faster. (I think the 65816 is faster clock per clock though, but I forget)


    He just seemed to rattle off a series of popular complaints with no personal knowledge of the Saturn to me. That is the kind of speculation I don't have the liberty of delving into in an objective review of a very subjective article though.
    Yeah, if he knew the Saturn architecture well, he'd likely be criticizing the poor documentation/tools as well as the high cost and some rather unnecessary features rather than claiming the 32x actually had any advantage in an apples to apples hardware sense. (anything the 32x/Genesis or even CD32x could do the Saturn could match and exceed in every way and with lesser difficulty -at least with similar documentation and tools)

    I didn't think his "colors per pixel" comment made any sense at all, and he definitely did not elaborate.
    He was implying the color depth (ie bits per pixel) was higher than anything else, but that's patently false. The maximum was 16 bpp using direct 15-bit RGB which is normal "highcolor" rendering and the same as what most 3DO, Saturn, PSX, and N64 games used (and some PC accelerators, though most used 6-5-5 RGB like SVGA rather than 5-5-5 RGB iirc). However, the 3DO, jaguar, PSX, and N64 all supported truecolor rendering (if rarely used) and the Jaguar supported the custom 16-bit CRY palette for super smooth shading comparable to truecolor but limited to a fixed set of 256 colors with 256 intensities (shades) output as 24-bit RGB. (SVGA was sporting highcolor and truecolor as well, though non-accelerated DOS games were sticking to 256 color for the most part)
    But the standard 32x color depth was 8bpp (256 colors) like MCGA/VGA from 1987 (games started appearing about 1989/90) and the SNES's highest color modes, and the SNES used 15-bit RGB like the 32x too. (VGA uses 18-bit RGB so 262,144 vs 32,768 colors)

    Software rendering is an option if the hardware manufacturer provides the info or the developer strains that relationship by reverse engineering. Obviously most developers are going to go ahead and use all resources available if that is an option.
    With the Saturn's lack of high-level support, Software rendering would be an option by default, and again, from what I understand several developers opted to lay off th quads and software render triangles for that very reason. (possibly the reason Doom is so slow if it really is a direct PSX port)
    Even with high-level emphasis you very well may have had a good amount of software rendering: obviously if "high level" simply meant a C compiler for the CPUs, but in the context of OS/API level programming is where it would be more iffy. (they managed it on the 3DO somehow with Doom... they didn't do that with polygons given the lack of texture warping -unless they did perspective correction like PSX doom)

    I would have to see significant primary evidence that the Saturn had the SH-2s added after its initial announcement in mid 1993. I can see a later stage DSP addition/modification, or even the VDP's and RAM being rethought, but not the SH-2s. The existence of the SH-1 in the Saturn only demonstrates Sega's willingness to stick to code compatible central processors in my opinion. This is similar to why the Sega CD uses a 68000 instead of using something arguably just as cost effective but incompatible with the Genesis CPU.
    Actually, the addition of the SH2s would be far more likely than any actual modification to the chipset, especially the claim that VDP1 added the 3D drawing modes or the DSP was added. (it was integrated into the SCU ASIC, so not slapped on the board at the last minute)
    Slapping a couple CPUs on the board would have been relatively simple, especially if the system was already designed around the architecture. (ie SH1)

    The use of the SH1 as the CD-ROM controller is crazy unless Sega had specifically planned it to be the CPU or a very prominent coprocessor when not managing the CD-ROM (ie when no loading occurred or CD-DA was playing, like the 68000 in the Sega CD).
    The SH1 in the Saturn can NOT be programmed by the programmer, it's a fixed function MCU given the comments from developers and homebrewers who've worked on the Saturn... unless that's just one more thing Sega never documented.
    In any case it's a waste that added a lot of unnecessary cost. (even a 68000 based MCU would be overkill, but at least more reasonable)
    If they were going to use the SH1 anywhere it would have made more sense as a flexible manager for the sound subsystem instead of the 68k like the ARM7 in the Dreamcast. (in fact they could have then gotten away with a considerably simpler sound set-up then, especially dropping the unused audio DSP... and given the general lack of use of the synthesizer capabilities, dropping to a plain DMA PCM ASIC would likely be preferable, or Yamaha's OPL4 for that matter -24 hardware PCM channels using 8, 12, or 16-bit PCM with up to 44.1 kHz plus the OPL3 FM block -SH1 would handle any compression or additional software mixing)

    I do think that SoJ made absolutely certain that the Saturn would have no deficiencies at 2D. No matter how arguable I find the Saturn's 3D prowess, I consider its 2D dominance that generation as close to fact as these things come.
    Actually it did. It had the amazing BG engine, but the "sprite" engine or blitter rather was relatively slow compared to the Jaguar or PSX (and in theory N64). So games pushing a lot of objects on screen, especially large objects, would eventually hit a wall on the Saturn VDP1's bandwidth and do so MUCH sooner than the PSX or Jaguar (PSX using texture rendering rather like the Saturn, but the Jaguar using its object processor which was only limited by the main bus bandwidth, and in fact is the only processor in the jaguar which can fully saturate the 106.4 MB/s bus bandwidth). So it's onl games that have very detailed BGs and use of many BG planes that the Saturn gains any real advantage.

    Well that and the added RAM, but that's still quite variable and at a big disadvantage for "sprites" due to the 512k for textures vs the PSX's shared 1 MB block. (744 kB for a 320x224 highcolor display) Though both can use main RAM for auxiliary storage. (the PSX has the advantage of a faster main bus too at 133 MB/s vs 114 MB/s for SDRAM and 1/2 the for DRAM in the Saturn -and much less than that for added RAM on the carts, albeit not nearly as bad as the Genesis's 3.8 MB/s cart bus -not sure about Saturn, but the jag has a 10.7 MB/s cart bus iirc)

    This always seems to be comparing the dual SH-2s to a computer system or workstation that would use dual CPUs to simultaneously work on separate code. Everything I've seen says that the Saturn was not designed for that purpose, instead the second SH-2 was included to work while the first was busy.
    Yes, the SH2s were not designed with that sort of parallel multiprocessing and in particular you can't have 2 sharing 1 bus and have both working in cache: only 1 can use the cache while the other only works in main RAM and/or scratchpad RAM, true for the 32x just as much as the Saturn. (but the 32x is worse off as you also have only a 16-bit bus) Hence the "slave" SH2 being used as a coprocessor. Even with 2 separate buses for the SH2s, you'd have to do a good amount of work to efficiently use both and wouldn't ever hit the performance of a single CPU at 2x the speed, but on a shared bus you'd be very lucky to hit 75% of that. (at least that's the impression of programmers and engineers who've dealt with the Saturn and 32x or similar systems)
    It's not quite as bad as the Jaguar where you have the slow 68k, GPU, and DSP all sharing one bus and no cache at all. (albeit if the CPU had a cache and the GPU and DSP were used as intended, that wouldn't be a major problem as most stuff should have been done in the small on-chip scratchpads... but with the cacheless 68k it's so bad that developers -especially homebrewers- found it far more efficient to use the GPU as the CPU instead, even though it's not very efficent as a CPU compared to it's rendering/lighting/transform/3D math capabilities as it was intended, and it's also got some bugs that hinder it considerably more for use as a CPU -the DSP is actually the same RISC core as the GPU but it's got a rather hacked/inefficent bus that makes it put a rather heavy penalty on bus accessing and is made worse by the 68k forcing it to use a 16-bit bus rather than the intended 32-bit one -let alone the cycle efficient 64-bit DMA of the GPU -the DSP was intended for audio operations that were primarily computation, so the bus issue wasn't significant)

    I consider this the same idea, though of course not the exact same implementation, to how modern PCs use dual core processors even when software is not optimized for them. Even though the real time rendering performance would improve in audio/video/3D editors if the software was dual core optimized, there is still a significant performance increase (nearly double in most cases) from a single core setup running at the same speed.
    No, it's very different as modern multi-core CPUs have parallel caching, L1 AND L2 caching, and not to mention the advent of DDR/DDR2/DDR3 etc with 128-bit DMA (dual channel) greatly facilitating multiprocessing without terrible I/O binding. (even with good caching there was a lot of I/O binding problems were still serious with many multi-CPU workstations prior to DDR -or RDRAM perhaps -not positive about the latter, though the bus sharing/IO limitations really depends on the circumstances -in cases where it's computation speed that limits things then of course bandwidth/bus sharing isn't going to be the issue)
    That's also a big reason the Xbox and 360 were able to use a single unified bus efficiently. (the N64 focused on that too with some trade-offs, albiet nowhere as severe as the Jaguar suffered -again mainly due to lack of a CPU cache)

    I agree, except that I think Sony basically created this problem by marketing high level development kits to third parties in the first place. Revolutionaries at Sony goes to great lengths to basically make it look as though Sony was "smart" and everybody else was holding the industry back for just that reason.
    No, Sony followed a trend that took a long time to reach common mass-market game consoles. Many PC games had already made the shift, though a fair amount of late DOS games had significant amounts of assembly language optimiation too, but there's a reason even code heads like Carmak were already pushing C in early 1994 and why some late Genesis games were even being programmed in C (often running into problems due to code density being too low and not fitting into the limited RAM -requiring aid from experienced assembly language programmers). Carmak was using a ton of C in his work on the Jaguar even and was working on a custom compiler optimized for the system. (Atari's compiler was complete crap and not really usable at all)

    The advent of windows and various APIs on PCs (for software rendering AND hardware acceleration) be it Glide, Rage, DirectX, OpenGl, etc that went a step beyond simple high-level programming into full OS and API level programming. (high level languages imply just that: BASIC, Fortran, C, C++, etc, etc compiled code, not direct assembler or machine code, but also far from an actual API)
    They also didn't make the mistake of 3DO with such a closed architecture, but given complaints it seems that they tools were a bit limited in terms of low-level documentation. (for early PSX dev kits)

    Sony was going with the flow of common mass-market trends that had been building for several years. The Jaguar was a 1989/90 design (when it was laid down on paper) that didn't have the benefit of such foresight for such features, let alone the funds to help address that late in design or after the fact with software solutions.
    The N64 was rather well oriented to high-level as well, if a bit constrained in other respects. (and programming custom RSP microcode is pretty much as low level as you can get, especially givne the basic tools SGI provided -albeit that was optional and not even a promoted aspect of the hardware -hence the use of the "standard" SGI microcode even with the alternate turbo 3D code available)

    The early software does_not demonstrate this rumor. Exclusive PS1 games seem to demonstrate the system's relative 3D prowess, but even they have exclusive Saturn counterparts (High Velocity, VFR/2, Cyber Speedway, WingArms) that gray that area completely.
    You'd really need the man hours, budget breakdown, and general development times to say that one way or the other.

    This alone discredits the idea that Sega was seriously considering such a solution after 1992. Something capable of porting Daytona and Virtua Cop (at a minimum) would have been on the table as soon as their arcade counterparts started development.
    The hype surrounding the 3DO and Jaguar if nothing else should have been major wakeup calls, so indeed a redesign MAY have taken place, but much earlier than other claimed. (and the specs specifying the use of the SH1 might have been based on early 1993/late 1992 figures that got leaked)

    I keep seeing the DSP referred to as some sort of helper chip for 3D, but the docs point to it being inserted as a special code processor, especially for DMA. I think there is some disparity as to whether the DSP was poorly documented versus Sega providing every example of special processes the chip could handle. I also don't trust the idea that the DSP was primarily for 3D calculations, as opposed other smaller processes involved in total system management. One of Sega's 1995 documents that I have seen claims that is exactly what the DSP was for.
    Note there are 2 DSPs in the Saturn, of course, 1 for audio and 1 for geometry, but the geometry DSP is on-chip with the System Control Unit which is a DMA controller. They're totally separate components on the same chip, it's an ASIC. (just like the 2 VDPs are still discrete in the ASIC implementation of the revised Saturn or all the consolidation that gradually occurred with the Genesis VDP ASIC, or hell, the Jaguar's 750k transistor massive system on a chip TOM with the GPU RISC, MMU, Blitter, and Object processor plus 4 kB of scratchpad RAM in one big ASIC -and very advanced in 1993 with .5 micron manufacting on top of all that -it was the largest ASIC that Toshiba had manufactured up to that point and the largest SoC in the world iirc)

    Could the Genesis alone have loaded data off the CD and processed it without *any* hit to load times?
    With the CD-ROM cache, yes, and much shorter load times obviously with less RAM to load into. (with the amount of RAM the PCE CD used max load times would likely be 1/8 that of the Sega CD)
    Aside from caching one major load time savings is use of redundant data... seeking kills load times in any case. (though that's the main utility of the buffer)
    The program RAM has nothing to do with load times, just how much memory can be loaded at one time, so by definition LONGER load times but less frequent ones. (as well as animation/detail that wouldn't be at all possible with less RAM)
    The 68k in the CD wouldn't at all be necessary either for that matter, though it does make a useful coprocessor. With no sub-CPU at all, the Gensis CPU would take the overhead of managing the CD drive which would significantly complicate on-the-fly loading and especially decompression, but a simple/cheap MCU could have offloaded that as well. (the 68k was obviously added for general purpose coprocessing -and gave a lot of untapped potential for software rendering)

    I'm not sure whether the X'Eye has more cache, I recall its load times not being significantly faster anyway. The X'Eye was supposed to have a much more optimized CD-ROM drive to eliminate "data access time." Mine certainly never impressed me over my Sega CD Model 1, I haven't spent much time with my SCD2.
    That could imply faster seek times or a larger cache, likely not a 2x speed drive though. (I think that might have screwed up compatibility)

    NEC interviews around the time of the Sega CD's development were almost exclusively promoting newer compression solutions for the PCE/TG-CD. My only experience with the PCE-CD experiencing frequent or long load times is with Altered Beast CD. It was a port of the 4mbit PCE HuCard with enhanced audio and some other (significant in my opinion) graphical enhancements. The game loaded though, about three to four times per level, totally freezing the action. NEC Avenue, for whatever reason, couldn't eliminate these loads in early PCE-CD games. Apparently they simultaneously worked on more advanced compression solutions and the System Card updates.
    Compression is one thing, use of redundant data for linear loading is another, both being software solutions more or less: a cache is a hardware solution and the only reason load times would be fundamentally shorter on the MCD. (due to the much larger amount of RAM alone, the MCD would imply much longer load times, but less frequent if not for the cache or various software solutions)
    One thing about compression was that the packed pixels of the MD graphics compressed far better than planar graphics using conventional lossless schemes. (SNES and PCE/CD developers eventually opted for using chunky pixel graphics that were converted to planar graphics after decompression)

    These were by far not foregone conclusions at the time.
    True, the Jaguar had only test marketed in 1993 and 1994 was the real launch year. But by the time sales figures started coming in, it should have been a wake-up call. (likely by mid 1994, but certainly after the holiday sales showed up -they weren't even on the map with well under 100k sold in the first year and only 135k sold by early 1996)
    The 3DO OTOH would have been more of a concern if the sales figures are accurate: especially given that was in spite of a huge price. (so a big warning if the price dropped)

    Welcome to the wonderful world of investment marketing. Since all of these are US sources, and we're speaking of 1994, I would suppose every comment is US centric. At the least they only make special mention of the world markets (still plural at the time) in parenthetical statements.
    The market itself may have been declining, but if you look at market share, Sega was leading at 55% for 1994, at least given the variety of sources claiming that mentioned here: http://sega-16.com/forum/showthread.php?p=304454
    Given that the "slump" never fully turned around until the 5t gen was in full swing in 1997, it would seem that market share would have been more of a concern in general as revenue was going to decline and/or plateau until the next-gen hit full-force. (at very least in hind sight)
    But really, the market analyst comments seem to mirror some of those claiming the NES was saturated and a "dead end" in 1990 when it had a very strong presence for several more years, but 1990 had simply been the peak (or '89 rather) followed by a gradual decline into the budget market enjoyed by all market leading game consoles. (2600, NES, Genesis, SNES, PSX, PS2, etc)

    Nintendo and their major third party supporters dominated the "million seller" lists for that generation.
    Yes, but a few big hits doesn't mean Sega wasn't outselling them with a broader array of software, just way more distributed. Again, this came up here:
    http://sega-16.com/forum/showthread....454#post304454
    In fact, having more compelling games rather than milking a few would split the market more in general as most people could only afford a few games a year, so there's no way they could push for all of them.

    I would put brand recognition as the prime candidate for why Sega and their major third party supporters didn't see higher sales on their original IP.
    How about in Europe? Plus, but 1993 Sega's name was as big on the mass market as Nintendo, at least to the same extent as Nintedo vs Atari in 1989. (ie you had average non-gamers who might have almost no knowledge of Nintendo but still remember Atari, though '87 and '88 were far more extreme)
    Of course, Sega screwed up that brand name with the horrible mistakes following 1994 (some in 1994, but the wost in 1995-1997)

    I suspect that if we could compare worldwide Genesis/MD sales to SNES/SFami sales going into 1994 the situation would look dire for Sega. We've already seen their financial situation, which alone mandated radical action.
    Why would it look dire? Japan was the only major market where Nintendo had a definitive lead (they were also burried under NEC)... otherwise the Genesis/MD had a significantly larger userbase overall and strong software sales. (per unit at least, revenue is another matter entirely)
    Albeit there seems to be a LOT of conflicting info on this, though Kent's article makes no specific reference to unit sales at all, but only revenue and even has a confused remake about "earnings" (ie revenue) coupled with profits that makes rather little sense.

    Cash cows are just that. That is why Sega focused on new hardware and software and Nintendo focused on software.
    Yes, and the Genesis was a cache cow that could have led on into 1997 at least before falling fully into the budget market and should have continued to be pushed though 1999 given Majesco's full 2 million in sales and significant software revenue as well. (mainly re-releases at that) That's several hundred milllion in revenue that Sega missed out on. (and could have been more had Sega been th one pushing it from 1995-1997 rather than pulling back by late 1995 and halting production until Majesco's offer)

    Bayless paints this as an assumption on SoA's part. The entire company in the US thought the 32X was going to be the only thing on the market until 1995. McFerran also states that SoJ had a "consensus" that Saturn should be emphasized above all else.
    Well then that's both Nakayma's fault for not providing enough information and/or Kalinske's for misinterpreting that.

    What do you think Kalinske would/could have done with project Mars if he knew full well that the Saturn (barring delays) WAS definitely going to be released in the fall of 1994 in Japan? (that's the main context where "jupiter" would have made sense, but the 32x and SoJ's mars would have made less sense)

    According to this interview it was a lack of communication on Nakayama's part exclusively. As a result of his actions the whole of SoA got excited about the 32X while SoJ provided exhaustive support and focused on getting the Saturn out by the end of 1994.
    OK, so that definitely seems like a big mistake on Nakayma's part and something SoA couldn't be faulted for.

    Most of my comments prior to this were assuming that SoA at least knew that the Saturn was going to be out in Japan by late 1994, but the western date was going to be at least a year after the 32x was released. (ie the context where the Jupiter or pure Genesis/CD and/or SVP/enhanced carts would have been better as a VERY short-term interim option intended primarily for 1994 and 1995)

    Everything I have seen shows Sega pointing to units sold through to consumers. This was even a major point of contention when the Dreamcast launched. When Sega reported one million units sold in less than two months people familiar with Sony dismissed it as production numbers not sales. But even EA knew they had sold that many units to consumers by November, and shifted their PR stance as to why they weren't going to support it.
    EA's stance was complete and utter BS of course.

    Not necessarily, especially not if marketed correctly. The Sega CD was already "bridging the gap" between high end PC games and 3DO games. It allowed Genesis owners a relatively low cost solution to play games they otherwise could not without significant investment. Somebody privileged with a mid range PC at the time would have a hard time seeing that.
    There's a LOT more the CD and Genesis alone (or with cheaper enhancement via an SVP cart) could have addressed though, especially with improved compression for FMV allowing video quality about 3x better than 1992. (and they could have pushed further with better dithering techniques and some careful use of shadow without putting a hurt on VDP bandwidth)

    With that said, I do not know of a single example of a company releasing forward compatible hardware. I especially cannot figure how a cart based "Saturn" with no backward compatibility would have benefited anything.
    You mean a console manufactuerer obviously as it happens with computers all the time (a lower-end version).
    And the reason is quite obvious: cost.
    They wanted a machine that was a fairly reasonable mid-range cost for 1994 and 1995 (along with serving as a lower-end alternative to the Saturn in parallel), and such a system could have been $250 when the 32x launched (ie about the same price as a Genesis+CD or Genesis+32x alone and the same price as the Jaguar) while sporting very impressive hardware that could even remain competitive though the generation, but also allow full upgradability at ever decreasing cost. (especially from the perspective of the Saturn which -if not for Sony- likely would not have dropped to $300 until 1996, not below $300 until the end of 1996 if not 1997 and possibly not dropped below $200 until 1998 -Sony forced both Sega and Nintedno's hands, but Nintendo had much cheaper hardware that meant it wasn't eating big losses)

    The Saturn at $400-450 would obviously have been totally impractical for 1994 and 1995 for that matter.

    And you yourself mentioned the greater ease and lower cost of developing cart based games. (I still maintain that similarly simple CD based games would be significantly cheaper -any added development costs would be heavily outweighed by reduced manufacturing risks and costs -and high margins)

    Sales heals all wounds. If there's a will there's an engine.
    True, but that's assuming the PSX didn't attract developers away too quickly.

    So I take it you think the Scavenger team demo lied that all of those demos were running 100% real time on a 32X? Do you have any sources for this? I have found Scavenger to be one of the more reputable developers from the time, and very hard core into creating extremely unique engines.
    you mean the Zyrinx demo? (I'm not sure of anything else)
    You need to realize that Demos, and real-world performance are nowhere near close to each other.
    If the Zyrinx demo didn't use the 68k then there should have been a chuk of resource for a fairly decent game with rather weak sound, but if it did use the 68k too, any game would be significantly slower.
    So yes, the on-screen poly counts likely wouldn't be a huge difference from Virtua Racing, VF, etc and likely noticeably less with texture mapping or smooth shading. (albeit you can do more with g-shading and/or textures then with a few more polygons... except for textures read from ROM will slow things a lot in many cases, so another attractive reason for g-shading)
    I wouldn't be surprised if Zyrinx crammed as many textures as they could in SDRAM, if not all textures in some cases.


    For me? Definitely, I was an arcade game whore. After Burner was my "Super Mario Bros", I wouldn't be a gamer without it. Shadow Squadron and Virtua Racing were my passionate obsessions in 1995 until I started playing Saturn games that Fall. I played Genesis and Sega CD games to be sure, and kept track of SNES games at friends houses (and played 3DO and Jaguar in kiosks) but I was more than entertained by 32X software.
    And what if games of similar quality had been on SVP or Sega CD, or a hypothetical SVP cart coupled with the CD? (assuming polygon counts and speed were fairly similar, but color was the only limit)

    The arcade ports on the 32x (sans the 3D) should have been on the CD by 1993 at the latest with more following in 1994 (and some should have been in 1992), namely Space Harrier, After Burner, Out Run, and perhaps Thunder Blade and Galaxy Force.
    The Sega CD should have been capable of SuperFX level 3D (with better texture support) and should have been able to manage Doom at least as well as the SNES (more compromises with color, bu perhaps better in some ways -and with all levels in any case, but necessary cuts)

    SVP could have addressed most 3D games to a reasonable extent and SVP combined with Sega CD should have managed polygon rates much closer to the 32x. (there might have even been the possibility of getting double the color using full-frame Shadow if a small screen size was acceptable -like the size of SNES Doom or Dirt Trax FX -that would apply to any software rendered game on the genesis and also FMV -except FMV could also use additional tile by tile palette optimization)
    6 days older than SEGA Genesis
    -------------
    Quote Originally Posted by evilevoix View Post
    Dude it’s the bios that marries the 16 bit and the 8 bit that makes it 24 bit. If SNK released their double speed bios revision SNK would have had the world’s first 48 bit machine, IDK how you keep ignoring this.
    Quote Originally Posted by evilevoix View Post
    the PCE, that system has no extra silicone for music, how many resources are used to make music and it has less sprites than the MD on screen at once but a larger sprite area?

  10. #10
    I remain nonsequitur Shining Hero sheath's Avatar
    Join Date
    Jul 2010
    Location
    Texas
    Age
    42
    Posts
    13,313
    Rep Power
    130

    Default

    Quote Originally Posted by kool kitty89 View Post
    Interesting: is that over 256 colors counting the score bar, or just the gameplay screen? I's imagine the full screen as otherwise it would imply the game was rendered in highcolor which would make little sense given that it's not clipped to 204 lines or less.
    The status bar is the only thing in-game rendered by the Genesis.
    Score bar, main character, and other human characters on screen average 220-260. One shot I have is 264 colors with the above plus a dithered grenade explosion. I didn't know about Irfanview when I measured these shots last, and Gif Movie Gear only goes to 256 colors, hence the greater than symbol.

    Quote Originally Posted by kool kitty89 View Post
    The SNES version obviously has to work around the 256x224 resolution as well as the 15 colors per tile/sprite limitation plus the 8 palettes for sprites and BG. (ie 8x15 color palettes for the sprite and another set for BG all selected from 15-bit RGB -same as the 32x) So a lot better than the Genesis, but limited compared to a 256 color bitmap display like VGA or the 32x. The SNES could have used mode 3 with 256 colors per tile, but that would use up a lot more memory and also hinder animation due to increased VRAM DMA time needed for updating.
    However all of that is a bit moot given the 32x used 3 times the ROM as the SNES (albeit a year later) with 3 full MB to the SNES's 1, so that's a massive difference right there for both audio and video.
    I've been told under no uncertain terms, by a respected member of this community, that more colors does_not impact VRAM, RAM or bandwidth in any way. The context of the discussion was an attempt to qualify Sega choosing not to put more CRAM in the Genesis/MD when NEC had more onscreen palettes in 1987. I figured that more on screen colors would affect something in regard to resources and was told unequivocally that because the tiles were still 16-colors each this was not the case.

    Quote Originally Posted by kool kitty89 View Post
    As for Sound, the 32x version has pretty noisy PWM and fairly low quality samples I think: I wouldn't say it sounds better than the SNES overall, but it's debatable to a degree. The 32x's sound system is more limited than the SNES's for sure, but in particular it's very limited by any officially released game due to the lack of working DMA in the dev kits, so ALL the PCM/PWM playback had to be handled in software using a ton of resources. (otherwise the only resource needed would be for software mixing multiple channels -again DMA and no DMA is the difference from the slave SH2 managing 32-64 channels vs 4-8 channels at 22 kHz)
    Once I get my SNES everdrive in I'll do a full comparison video. I don't recall any noise on the 32X game on real hardware. If you're using emulation are you definitely not using any filters on either version?

    Quote Originally Posted by kool kitty89 View Post
    Not used more, but fundamentally better and FAR more cost effective. There is a middle ground and that would be a powerful and flexible coprocessor that's still more focused than a CPU but a lot more flexible than a dedicated blitter or GPU (like a DSP or the RISC "GPU" in the Jaguar), but really it's probably most efficient to do a bit of both.
    In the case of the PSX it was fully dedicated hardware with the main flexibility being the CPU while the GTE was solely for geometry calculations and the GPU was only really useful as a blitter/rasterizer: ... Interestingly enough, the GTE in the PSX was one area that PC graphics cards took a long time to push in hardware as most early 3D cards ... required the CPU to handle the 3D math while the accelerator handled the rasterization/drawing/textures/shading/filtering/perspective correction/etc. ...
    Right, and I think the PS1's soaring popularity after 1997 subsequently mandating near simultaneous PC/PS1 software releases had something to do with that. In my mind, Matrox Millennium/Mystique cards represented a more Saturn-like approach (not sure about whether they use quads though). Matrox cards were way ahead of their time in texture mapping capabilities at least. I still wish I could get the Mystique optimized Destruction Derby 2 and Mech Warrior 2 running on a current PC. Only after the craze for "hardware accelerated 3D" and on board transform and lighting became the norm did texture map resolution/color depth catch up. The Saturn and Matrox cards had ruined me.

    Quote Originally Posted by kool kitty89 View Post
    As to polygons: trips vs quads is a whole other issue too: in the real world triangles just make a lot more sense and they did 20 years ago or 30 years ago too, or longer. The algorithms and general math used and developed by humans is tied very heavily to triangles such that pushing quads really complicates that among other things.
    Trinitarians would love that statement, if it is indeed universal fact that is.

    Quote Originally Posted by kool kitty89 View Post
    Then there's the general flexibility of triangles vs limitations of using quads: you can easily pair triangles for quadrilateral surfaces (let alone larger polygons), but you can't warp a quad into a polygon (ie overlap 2 points) unless it's solid shaded: doing so with a gouraud shaded or texture mapped quad will horribly warp the texture or screw-up the shading.
    I can see this with more complex environments/models, but not with 90s 3D. More below.

    Quote Originally Posted by kool kitty89 View Post
    There's a reason pretty much all (polygonal) 3D games prior to hardware acceleration used triangle rendering (including the 32x and jaguar) and also a reason 3D workstations and early 3D arcade boards predominantly did as well. ... The early PC 3D accelerators also pushed polygons with the solitary exception of NVidia's NV-1 and canceled NV-2 chipsets. ...
    The first time I saw an SGI workstation the only 3D object the artist showed me was a dog house. It was an object comprised of a bunch of squares and two triangles. I'm not going to take a stand against triangles, I think everybody ought to have some, but back then they just weren't the only polygon in town.

    Quote Originally Posted by kool kitty89 View Post
    Sony wasn't setting a trend, but solidifying it: the bigger thing that the 3DO, Saturn, PSX, etc were pushing was polygonal 3D over the main competition on software rendering consoles (namely PCs) ray-casting based height mapping (including Wolf 3D, Doom, Duke 3D, and voxel engines). There was every possibility of hardware acceleration aiding ray-casting as well, but it didn't happen that way.
    Now, I do have to say, having learned the Doom engine well enough to design my own maps back in the day, the engine had severe 3D limitations if it was 3D at all. Not being able let one "floor" overlap another in 3D space was extremely limiting and took up a ton of space in the maps. If that was a limitation intrinsic to that type of pseudo 3D it was an idea as doomed as quad engines.

    Quote Originally Posted by kool kitty89 View Post
    As it was, it's almost certain that a major reason the Saturn and 3DO used quads was purely due to engineering time/cost due to the rectangle based "sprite" rasterizer/blitter (like what the Sega CD has) being just a step below a 4-point warped quad rasterizer while a triangle rasterizer would take a fair bit more work.
    That more than implies that there was some sort of resource trade off. Whether it be the "newness" of triangle rendering or not, something compelled very notable engineering firms to go with quads instead of triangles.

    Quote Originally Posted by kool kitty89 View Post
    Yes, but any such games would look crappy compared to what the Saturn was capable of.
    ROM limitations wouldn't have caused the same thing with "Jupiter"?

    Quote Originally Posted by kool kitty89 View Post
    It should have been fairly easy to port some 32x games to Saturn, but applying significant enhancement would be another story: the simplest thing would be speed improvements from the fast main RAM compared to very slow ROM on the 32x and SDRAM 1/2 the bandwidth of the Saturn and no need for on-the fly decompression, plus the ~25% faster CPUs. But even for a straight port you had the Genesis hardware complicating things and the sound engine would need to be completely re-done. ... Not to mention games running the game logic/AI on the genesis CPU. ... Any polygonal games would be pretty stuck with software rendering...
    I think this understates the similarities in the ways the Genesis-32X and the Saturn's VDPs were designed to render 3D (3D overlayed on 2D). Consider a scenario where the Saturn was going to be in Japan only until late 1995 at the least and the 32X saw full second and third generation software. The ease of down-porting Japanese Saturn software to the 32X and up-porting western developed 32X games to the Saturn would have been greater than adapting either to other platforms. After 1992 we already saw a good number of third parties (especially the crappy ones) developing games concurrently for Genesis, SNES and Sega CD.

    Could part of the "Mars"-Saturn scenario have been intentionally making it difficult to port games to non-Sega hardware?

    Quote Originally Posted by kool kitty89 View Post
    Then there's the fact that you might end up wasting the slave SH2 in the Saturn if a 32x game dedicated the slave SH2 to handling audio ... , so the game would need to be modfied to take advantage of the added CPU as well. Then you'd also have highcolor vs 256 color 32x rendering on the Saturn, unless you pulled back to the 256 color mode.
    If more is always better, this should not have been a problem.

    Quote Originally Posted by kool kitty89 View Post
    So some 32x games could have simple/sloppy ports to the Saturn, especially with no or minimal use of the Genesis hardware, but they wouldn't usually be a ton better without further optimization.
    They probably should have done that for the better 32x games for sure and it's pretty odd that they didn't: especially Doom ...
    I'm almost certain that Doom's limited PS1 exclusivity was an early example of Sony's heavy-handed anti-competitive philosophy.

    Quote Originally Posted by kool kitty89 View Post
    Then there's VIrtua Racing ..., Cosmic Carnage, Metal Head ..., Star Wars Arcade ..., and you had some others, but I'm not sure how many were even out only he 32x by mid 1995.
    Virtua Racing was only a month late to the Saturn's early US launch and is essentially a more advanced version of VR Deluxe. Then there's Galaxy Fight, Ghen War, and Wing Arms or Panzer Dragoon (Star Wars Arcade is a very small step above a rail shooter). Unless the topic of the discussion is the importance of 32X software/brands for the early success of the Saturn, I don't see the benefit of adding these games to the 1995 Saturn library aside from splitting their sales up even more than in reality..

    Quote Originally Posted by kool kitty89 View Post
    ... They wouldn't have merited the Saturn being launched in May, but would have been nice additions nonetheless, especially alongside Kolibri, Tempo, Chaotix, etc or for a more moderate early launch like August (3rd party games are more dependent on getting interest in Saturn development early on).
    Third party support for the Saturn was in fact higher than PS1 support in 1995. For some reason I think more 2D or psuedo 3D platformers in the early Saturn library would have been great though.

    Quote Originally Posted by kool kitty89 View Post
    Unless Sega could have ramped up Saturn production and collectively planned the may launch with a dozen plus games, a healthy number of systems, developers and retailers notified months ahead of time, western Saturn development starting by fall of 1994, etc. (some of that dependent on SoJ, but otherwise the only remaining drawback would be the $400 price tag, unless they were willing to dump that to $300 or perhaps $350)
    All of that is just trying to make the Saturn a mass market product earlier than it was ever going to be. Even the PS1 wasn't mass market that year.

    Quote Originally Posted by kool kitty89 View Post
    But a major point would be that the Saturn would not have helped developers transition to the Saturn in general: it would have promoted tendencies that would often do nothing but not promote the Saturn's best performance. The specifics and quirks of the architecture in the Saturn, especially making good use of the VDPs. (the main thing the 32x would not do at all)
    You mean tendencies like assembly level software development for multiple CPUs, VDPs and sound processor capabilities and 3D overlayed on 2D?

    Quote Originally Posted by kool kitty89 View Post
    Yes, but why didn't SoA know the Saturn was going to launch then?
    In that context they would have either canceled the 32x or tried hard to put a positive spin on the Saturn.
    HAD the Saturn been held back, the 32x definitely would have made more sense than it did, maybe not necessary, but more reasonable. (especially if the Saturn was held to allow more modifications, refinement, optimization, and especially put together good development kits)
    Based on Bayless' suspicions I would say that Sega of Japan didn't know whether the Saturn would make the 1994 launch window. Why, or whether, they felt that way could have had to do with any number of organizational, development, contractual, marketing, software or financial reasons.

    Quote Originally Posted by kool kitty89 View Post
    That sounds like a prime example of miscomunication due to there being every indication of SoJ wanting the Saturn out ASAP and aiming at a 1994 release, especially to meet (or preferably beat) Sony in the market. (you mention Nakayma was worried about the Saturn being ready for 1994, but did he ever tell SoA that they were or were not planning to release it in 1994?)
    Or a classic example of human beings misjudging the future and then failing to adequately correct said mistep in the present. There is no indication that SoA was not aware of the Saturn's launch by mid year at least. The facts point to them being nearly unanimously disappointed in that knowledge though. That disappointment may have caused any number of problems regarding PR, software, or even final stage 32X design optimizations.

    The surprise of the Saturn's '94 Japanese launch almost definitely screwed up SoA's culture and subsequently their public relations and marketing. There was no reason at the time to have Japanese console envy, most consoles were launched a year earlier in Japan. Sega as a company was well versed in supporting multiple generations of hardware simultaneously. So, I can only attribute SoA's mental break down at the Saturn's launch to a breach in trust between the divisions or some larger (revenue oriented) issue.

    Quote Originally Posted by kool kitty89 View Post
    Mainstream is the definition of ignorant consumer... at least if Xbox 360 sales are any indication.
    And yet they make the industry go round, right down to what kinds of games are created and how marketing dollars are spent. I feel the need to correct popular misconceptions mostly because I am so bored with the state of the industry today.

    Quote Originally Posted by kool kitty89 View Post
    But on the most fundamental level: 3D in computers and video games implies the use of points plotted to 3D space (the main reason ray-casting isn't usually considered real 3D), somthign quad and triangle engines do, but somthign a fully "2D" game could also do. (ie you could have a fully scaling/animated or scaling/rotation "sprite" based game that actually maps out the objects in 3D space and then renders the "sprites" accordingly -I'm not sure if Wing Commander does that, but that's sort of the idea)
    I would have clipped this, but I decided to take the opportunity to tell you that I agree with the above completely. I have often been disappointed in "fully 3D" games that might have been done just as well or better with sprite scaler engines.

    Quote Originally Posted by kool kitty89 View Post
    No, not purely lines of code, but cycle times: the 65815 has very fast instructions compared to the 68k, but also simpler instructions while the 68k's are more powerful, so it depends on the case: for operations that cater well to the simple instructions the 650x arch has a huge clock/clock advantage over the 68k, but for stuff that makes use of the more powerful instructions of the 68k, it can very well have an advantage too. ... The same is true for the PCE's CPU, albeit it's considerably faster. ...
    Right, I just haven't seen any discussion on the topic reach a definitive conclusion. Also, the software evidence leans absolutely towards the Genesis being superior in all things but colors and audio. I think this could be attributable to specific capabilities of the Genesis VDP, but it is more commonly attributed to relative CPU prowess in regard to (MD specific) games.

    Quote Originally Posted by kool kitty89 View Post
    Yeah, if he knew the Saturn architecture well, he'd likely be criticizing the poor documentation/tools as well as the high cost and some rather unnecessary features rather than claiming the 32x actually had any advantage in an apples to apples hardware sense. (anything the 32x/Genesis or even CD32x could do the Saturn could match and exceed in every way and with lesser difficulty -at least with similar documentation and tools)
    His opinion reflects the common perception in the industry though, and especially late 90s early 2000s conjecture. I doubt Bayless has spent any time arguing with homebrew developers and enthusiasts.

    Quote Originally Posted by kool kitty89 View Post
    He was implying the color depth (ie bits per pixel) was higher than anything else, but that's patently false. The maximum was 16 bpp using direct 15-bit RGB which is normal "highcolor" rendering and the same as what most 3DO, Saturn, PSX, and N64 games used ...
    But the standard 32x color depth was 8bpp (256 colors) like MCGA/VGA from 1987 (games started appearing about 1989/90) and the SNES's highest color modes, and the SNES used 15-bit RGB like the 32x too.
    That is a touch sloppy. The SNES could pull from a 32,768 master palette of colors, but could only display eight 15-color palettes simultaneously in the most common modes. I am not aware of the 2048 "Direct Color Mode" ever being used. The SNES could not display its master palette though, and the 32X apparently could, though that ability was equally impractical to the SNES' Direct Color Mode.

    Quote Originally Posted by kool kitty89 View Post
    Actually it did. It had the amazing BG engine, but the "sprite" engine or blitter rather was relatively slow compared to the Jaguar or PSX (and in theory N64). So games pushing a lot of objects on screen, especially large objects, would eventually hit a wall on the Saturn VDP1's bandwidth and do so MUCH sooner than the PSX or Jaguar .... So it's onl games that have very detailed BGs and use of many BG planes that the Saturn gains any real advantage.
    Somehow this does not show up in software.

    Quote Originally Posted by kool kitty89 View Post
    Well that and the added RAM, but that's still quite variable and at a big disadvantage for "sprites" due to the 512k for textures vs the PSX's shared 1 MB block. (744 kB for a 320x224 highcolor display) Though both can use main RAM for auxiliary storage.
    That's oversimplified again. The Saturn had a total of 1.5MB of VRAM to the PS1's 1MB. The PS1 had to render everything, including backgrounds and framebuffer, to that 1MB. The Saturn, as you noted, had backgrounds down pat without any hit to that 512 "texture RAM," but it also had 256KB framebuffers for both VDPs. Also, developer interviews and published software show the PS1 hitting a "wall" in 2D animation and color counts and the Saturn not hitting a limit.

    Quote Originally Posted by kool kitty89 View Post
    No, it's very different as modern multi-core CPUs have parallel caching, L1 AND L2 caching, and not to mention the advent of DDR/DDR2/DDR3 etc with 128-bit DMA (dual channel) greatly facilitating multiprocessing without terrible I/O binding...
    I am sure there are enumerable differences in implementation. I was referring to the overall design idea as opposed to your suggestion of a faster single CPU that likely cost more than double the dual CPU solution. Cramming half again the transistors while doubling the speed is usually much more expensive than two processors assisting one another and *not* fast enough to justify the cost difference.

    Quote Originally Posted by kool kitty89 View Post
    No, Sony followed a trend that took a long time to reach common mass-market game consoles.
    Well, the entire industry has either missed the topic of dev kits entirely, or fervently believe that Sony revolutionized the industry with them.

    Quote Originally Posted by kool kitty89 View Post
    The advent of windows and various APIs on PCs (for software rendering AND hardware acceleration) be it Glide, Rage, DirectX, OpenGl, etc that went a step beyond simple high-level programming into full OS and API level programming. (high level languages imply just that: BASIC, Fortran, C, C++, etc, etc compiled code, not direct assembler or machine code, but also far from an actual API)
    Yes, and Sega themselves teamed up with Microsoft for Direct X during the Saturn's days for Sega PC softs, and then again with WinCE on Dreamcast. I'm not arguing the trend itself, or even its merits (lower cost development was a plus for everybody involved).

    Quote Originally Posted by kool kitty89 View Post
    You'd really need the man hours, budget breakdown, and general development times to say that one way or the other.
    Since most of that will never be known, release dates are the only real indicator. The Saturn, with no high level dev kits, managed exclusive versions of almost all of the popular PS1 titles within a month or two (i.e. they weren't a reaction).

    Quote Originally Posted by kool kitty89 View Post
    The hype surrounding the 3DO and Jaguar if nothing else should have been major wakeup calls, so indeed a redesign MAY have taken place, but much earlier than other claimed. (and the specs specifying the use of the SH1 might have been based on early 1993/late 1992 figures that got leaked)
    This I would totally agree with. There were too many rumors in 1991 of a System 32 based console for it to be complete magazine fabrication. I suspect that, like the 32X and "Jupiter", the Sega CD's popular reception curbed that project. Alternately, Sega might have simplified the 2D aspects and started focusing on how to create a 3D home console capable of porting their arcade games.

    Quote Originally Posted by kool kitty89 View Post
    Note there are 2 DSPs in the Saturn, of course, 1 for audio and 1 for geometry, but the geometry DSP is on-chip with the System Control Unit which is a DMA controller. They're totally separate components on the same chip, it's an ASIC. (just like the 2 VDPs are still discrete in the ASIC implementation of the revised Saturn or all the consolidation that gradually occurred with the Genesis VDP ASIC, ...
    That is interesting. Integration to a single chip early on usually implies some sort of shared responsibility. I had not noticed a second digital signal processor in the Saturn previously mentioned though. So the Saturn had the SH-1, Motorola 68EC000, and a DSP apart from the dual SH-2s, Dual VDPs, SCU and geometry DSP. Not to mention a much higher product quality CD-ROM drive than the PS1's and more memory overall. Hot dang they went all out, I wonder if we will ever see notes on why every component was chosen.

    Quote Originally Posted by kool kitty89 View Post
    With the CD-ROM cache, yes, and much shorter load times obviously with less RAM to load into. (with the amount of RAM the PCE CD used max load times would likely be 1/8 that of the Sega CD)
    Aside from caching one major load time savings is use of redundant data... seeking kills load times in any case. (though that's the main utility of the buffer)
    Right, CD-ROM games with less RAM means shorter levels or more frequent loads. Thus to allow levels of comparable lengths to contemporary cartridge games the Sega CD needed more than just CD cache.

    Quote Originally Posted by kool kitty89 View Post
    The 68k in the CD wouldn't at all be necessary either for that matter, though it does make a useful coprocessor. With no sub-CPU at all, the Gensis CPU would take the overhead ...
    That was primarily where I was coming from. This is much the same as the SNES versus Genesis in-game performance discussion. We do not have pertinent facts regarding what percentage of the SNES CPU was being used in any production level software to say that it "could have done more."

    What would we have sacrificed in regard to what the Genesis CPU's in game responsibilities to manage loading efficiently? Just more time with a blank screen?

    Quote Originally Posted by kool kitty89 View Post
    Compression is one thing, use of redundant data for linear loading is another, both being software solutions more or less: a cache is a hardware solution and the only reason load times would be fundamentally shorter on the MCD. (due to the much larger amount of RAM alone, the MCD would imply much longer load times, but less frequent if not for the cache or various software solutions)
    One thing about compression was that the packed pixels of the MD graphics compressed far better than planar graphics using conventional lossless schemes. (SNES and PCE/CD developers eventually opted for using chunky pixel graphics that were converted to planar graphics after decompression)
    There they are again, packed pixels versus planar graphics. We're talking about tiles versus full screen frame updates? You are correct that the NEC compression comments emphasized more data per disk and not more data per level. But the context was "better games."

    Quote Originally Posted by kool kitty89 View Post
    Yes, but a few big hits doesn't mean Sega wasn't outselling them with a broader array of software, just way more distributed. Again, this came up here: ...
    It would be much more conducive to conversation if you would summarize the points and cite any sources used. As it stands, the facts I have seen indicate that Sega's games were generally commercial failures with few that drew consumer interest like NES franchises could. VGChartz (I know, I know) had an article a couple of years ago that claimed to have found total software published for each system, and Genesis software was still way behind. That article has long since vanished though.

    Quote Originally Posted by kool kitty89 View Post
    How about in Europe? Plus, but 1993 Sega's name was as big on the mass market as Nintendo, at least to the same extent as Nintedo vs Atari in 1989. (ie you had average non-gamers who might have almost no knowledge of Nintendo but still remember Atari, though '87 and '88 were far more extreme)
    Of course, Sega screwed up that brand name with the horrible mistakes following 1994 (some in 1994, but the wost in 1995-1997)
    I hate to hurt any feelings, but Europe's market size was relatively insignificant. Except in the sense that European developers focused on locally popular platforms, software or hardware sales there was barely a dent in the Jp/US market.

    Quote Originally Posted by kool kitty89 View Post
    Yes, and the Genesis was a cache cow that could have led on into 1997 at least ...
    Certainly, but only at a retail price similar to any individual cartridge game. This does not bode well for a revenue problem like Sega was facing for the previous four years.

    Quote Originally Posted by kool kitty89 View Post
    What do you think Kalinske would/could have done with project Mars if he knew full well that the Saturn (barring delays) WAS definitely going to be released in the fall of 1994 in Japan? (that's the main context where "jupiter" would have made sense, but the 32x and SoJ's mars would have made less sense)
    SoA could have prepared a much more concise marketing stance regarding the cost effectiveness of the 32X for existing Sega owners. As it was they more than implied that Saturn was out of reach both in price and in release date until the Saturn's Japanese launch. After that point the entire mood regarding Sega in general turned for the worse in the media especially.

    If SoA had known the Saturn was going to launch that year overseas, they could have easily painted a different picture like I have. If that recent arstechnica article about inflation and software article is correct, the Saturn and Playstation were at least 30% more expensive in modern dollars. It should not have been hard to market the 32X on a gameplay and cost conscious bent.

    Quote Originally Posted by kool kitty89 View Post
    OK, so that definitely seems like a big mistake on Nakayma's part and something SoA couldn't be faulted for.
    It probably was Nakayama's fault alone. Even this article alone describes him going against SoJ's "consensus" for the Saturn for the 32X's creation. Whatever he had in mind though, *something* caused Sega to launch the Saturn too early and that thing was not the PS1 alone.

    Quote Originally Posted by kool kitty89 View Post
    Most of my comments prior to this were assuming that SoA at least knew that the Saturn was going to be out in Japan by late 1994, but the western date was going to be at least a year after the 32x was released. (ie the context where the Jupiter or pure Genesis/CD and/or SVP/enhanced carts would have been better as a VERY short-term interim option intended primarily for 1994 and 1995)
    ...
    EA's stance was complete and utter BS of course.
    PR across the industry continually lost credibility from 1994 on. I don't believe in coincidences. Sega merely contradicted themselves with the 32X and Saturn, other companies were actively involved in public deception.

    Quote Originally Posted by kool kitty89 View Post
    And you yourself mentioned the greater ease and lower cost of developing cart based games.
    That is, in fact, the only reason I think Jupiter might have worked somewhat. But for me personally, knowing that I was buying the lesser hardware, I just would have saved up for the full version unless there was an incremental option of equal cost. So now we're discussing a Jupiter, Jupiter CD-ROM, and a Saturn. [/quote]

    Quote Originally Posted by kool kitty89 View Post
    you mean the Zyrinx demo? (I'm not sure of anything else)
    You need to realize that Demos, and real-world performance are nowhere near close to each other.
    If the Zyrinx demo didn't use the 68k then there should have been a chuk of resource for a fairly decent game with rather weak sound, but if it did use the 68k too, any game would be significantly slower.
    Zyrinx was only one Scavenger team, Lemon's demo at the end of the video is almost definitely showing part of AMOK's engine. It is also worthy of note that even within the relatively low quality video the framerate dips during the Zyrinx portion during the texture mapped gouraud section. Similarly, Among the Zyrinx demos there is very little panning or "strafing", so this might be indicative of framerate limitations.

    Quote Originally Posted by kool kitty89 View Post
    And what if games of similar quality had been on SVP or Sega CD, or a hypothetical SVP cart coupled with the CD? (assuming polygon counts and speed were fairly similar, but color was the only limit)

    The arcade ports on the 32x (sans the 3D) should have been on the CD by 1993 at the latest with more following in 1994 (and some should have been in 1992), namely Space Harrier, After Burner, Out Run, and perhaps Thunder Blade and Galaxy Force.
    The Sega CD should have been capable of SuperFX level 3D (with better texture support) and should have been able to manage Doom at least as well as the SNES (more compromises with color, bu perhaps better in some ways -and with all levels in any case, but necessary cuts)
    Sega just never marketed, and probably never envisioned, the Sega CD for all of these purposes. You have seen how particular actors get type cast for a singular role when they are capable of much more. Similarly, if the Unreal 3 engine had been demoed with a 2.5D platformer instead of an FPS and Gears of War it might not have ever been used for such diverse functions.

    I doubt that Sega failed to understand the Sega CD's full capabilities. So, I can only suppose that the Genesis radically superior hardware and software sales made the difference if the expense of 3D development did not.
    Last edited by sheath; 10-08-2010 at 01:16 PM.

  11. #11
    Hero of Algol kool kitty89's Avatar
    Join Date
    Mar 2009
    Location
    San Jose, CA
    Age
    30
    Posts
    9,724
    Rep Power
    63

    Default

    Quote Originally Posted by sheath View Post
    Score bar, main character, and other human characters on screen average 220-260. One shot I have is 264 colors with the above plus a dithered grenade explosion. I didn't know about Irfanview when I measured these shots last, and Gif Movie Gear only goes to 256 colors, hence the greater than symbol.
    OK, so 260 total colors, not just on the gameplay window, so not highcolor. (if it were highcolor, 260 would be really low ) The Genesis graphics bump it up a fair bit: if you were careful you could manage up to 316 colors without added tricks, 370 colors with careful use of shadow, and a fair bit more if palette-reloading and/or hilight were used. (all assuming no shared palette entries aside from transparent -in practical use you'd likely have quite a lot of common palette entries on the genesis side and maybe some common for genesis and 32x as well)
    I'd bet Kolibri also has over 256 colors.


    I've been told under no uncertain terms, by a respected member of this community, that more colors does_not impact VRAM, RAM or bandwidth in any way. The context of the discussion was an attempt to qualify Sega choosing not to put more CRAM in the Genesis/MD when NEC had more onscreen palettes in 1987. I figured that more on screen colors would affect something in regard to resources and was told unequivocally that because the tiles were still 16-colors each this was not the case.
    No, not, that's not what I mean: CRAM has nothing to do with it, and for the Genesis, color count does indeed have no impact (aside fromshadow which effectively doubles the bandwidth), but the PCE and SNES (and NES and SMS and Amiga and ST, EGA and others) use planar graphics, not chunky/packed like the Genesis/32x/VGA/later systems (and some older ones like the A8/5200 or CGA among others). Though had the Genesis had a full res 256 color mode, it would have used double the bandwidth too, like Shadow more or less does. (using 8bpp rather than 4, thus 2x the bandwidth for updates and 2x the VRAM space)

    With bitplanes you can literally vary the bit depth on a cell basis: ie a 16-color cell is 4 bitplanes but can be dropped to 8 colors with 3 bitplanes, 4 colors with 2, or 2 colors with 1 and correspondinly decrease the bandwidth and VRAM space used. (unlike lossless compression which only saves ROM space) In the case of the SNES's mode 3 you have up to 8 bitplanes in use, so double the VRAM space and bandwidth of full 4-bit tiles.
    If you use only 8 or 4 colors per tile on the Genesis it wouldn't help at all as they're packed pixels with fixed 4bit pixel values, not storing pixel data spread out among several bitplanes. (planar graphics are also a pain in the ass to software render with for that reason -writing to multiple bitplanes to modify one pixel, and even impose limitations on blitters for similar reasons, but those can be somewhat more optimized at least)

    And again, this is colors PER tile, or for a full bitmap display, not "on-screen colors" which is a very vague comparison anyway. Yes, increasing the subpalettes won't impact VRAM usage or bandwidth, but that still limits the palette entries per tile, even the Neo Geo and high-end arcade machines of the time were only 4bpp with 15 colors per tile, but the neo has 256 individual palettes vs the 16 in the SNES, and neo Geo games often don't put a whole lot more than 256 colors on screen at all due to the high degree of redundant color values. (it does save a lot of space for uncompressed graphics though, 4 bpp is 1/2 the space as 8bpp and 1/4 that of 16bpp, so as massive as the arcade ROMs were, they'd have been much larger had they not beeb limited to 15 colors per tile/sprite)

    That's critical for software rendering especially as you're managing a single palette bitmap display and are thus stuck with 15 colors for 4-bit tiles no matter how many palettes you have: hence The Genesis being as good or better at the polygon/scaling layer in Star Fox and Vortex, but more limited compared to the mode 3 games (Doom, Stunt Race, and Dirt Trax -hence the reduced dithering) I think some of those (if not all) migth actually cut to less than 8 bitplanes to reduce space: I suspect 6 bitplanes for Doom and Dirt Trax especially. (which is 64 colors, or 61 rather, per tile -as you still have the transparent values stuck in there I think, at least if you want to use sprites and/or 15-color overlay tiles)
    However, on the Genesis you COULD do something somewhat like that via Shadow, but that would only get you a 31 color bitmap display and effective 8bpp tiles (but the shadow colors do expand the palette beyond the 512 normally available -so better shading among other things). It might be preferable to dithered 15 color graphics depending on the case -especially if you wanted to use a double wide pixel renderer like Doom. (among others when desirable -doesn't save VRAM space/bandwidth, but does accelerate rendering)

    Does that all make sense?


    Once I get my SNES everdrive in I'll do a full comparison video. I don't recall any noise on the 32X game on real hardware. If you're using emulation are you definitely not using any filters on either version?
    PWM noise is considerably worse on real hardware: fusion elimiates the horrible squealing integral to PWM (unless you filter the crap out of it and ruin any high frequency stuff), but the static is there in either case, again worse on real hardware. (and more muffled due to filtering -Fusion's filtering helps too though, and wihout as much detriment)
    PWM is a very poor method for digitized audio playback, especially at lower sample rates... it's equally problematic on the GBA and DS, though max sample rate stuff helps. (you don't see 32x games pushing 44.1 kHz 9-bit audio though... and that's better than the GBA/DS's 32 kHz 9-bit max output... if DMA worked, software decompression would likely facilitate considerably higher sample rates)


    Right, and I think the PS1's soaring popularity after 1997 subsequently mandating near simultaneous PC/PS1 software releases had something to do with that. In my mind, Matrox Millennium/Mystique cards represented a more Saturn-like approach (not sure about whether they use quads though). Matrox cards were way ahead of their time in texture mapping capabilities at least. I still wish I could get the Mystique optimized Destruction Derby 2 and Mech Warrior 2 running on a current PC. Only after the craze for "hardware accelerated 3D" and on board transform and lighting became the norm did texture map resolution/color depth catch up. The Saturn and Matrox cards had ruined me.
    No, the NV-1 was the only one really like the 3DO and Saturn in terms of 3D. You cerainly have others with more 2D acceleration too, but that's nothing like the Saturn's VDP2 and more like a powerful blitter. (somewhat akin to what the PSX does, but aimed at windows acceleration)
    The Mystique is actually more like the Jaguar in some respects, especially the focus on gouraud shading over texture mapping. (the Jaguar's object processor might haev actually meshed fairly well as a windows accelerator -optimized at moving around large numbers of rectangles at variable sizes at high speed -and also zooming/scaling)

    Note the Rage 3D cards also integrated 2D, the Voodoo didn't as it was a high-end add-on intended to be used like the 32x. (taking analog RGB from the 2D card and mixing with its own VGA output)

    But again, I was talking about the point/vetrex calculation which was done by the CPU (normally using floating point calcuations) as with software rendered 3D (Quake required an FPU), but with the actual rendering offloaded the the accelerator. (rasterization, shading, texture mapping, often Z-buffering filtering and perspective correction as well) In those cases they were again floating point math, not fixed point as the PSX's GTE used and SVP and Super FX and any CPU/DSP plotting done without an FPU. (ie 32x and Saturn were pretty much all fixed-point math as well, let alone software rendered Genesis polygon games -and amiga and St, older DOS games, etc)

    Trinitarians would love that statement, if it is indeed universal fact that is.
    Did you read the thread I posted?


    I can see this with more complex environments/models, but not with 90s 3D. More below.
    Tons of 3D models used lots of triangles... but even quadrilaterals are easier to deal with as triangle fans/pairs, let alone other polygons. Even when you do use a quad rasterizer, you'd be using triangle based math to calculate the points.



    The first time I saw an SGI workstation the only 3D object the artist showed me was a dog house. It was an object comprised of a bunch of squares and two triangles. I'm not going to take a stand against triangles, I think everybody ought to have some, but back then they just weren't the only polygon in town.
    That doesn't mean it wasn't working fully with triangles.
    Sega's model 1 games are full of quad-heavy models, but I'm almost positive all that is doen with triangles. (and it's a real shame they didn't use more triangles as quads have an uglier faceted look the way they used them)



    Now, I do have to say, having learned the Doom engine well enough to design my own maps back in the day, the engine had severe 3D limitations if it was 3D at all. Not being able let one "floor" overlap another in 3D space was extremely limiting and took up a ton of space in the maps. If that was a limitation intrinsic to that type of pseudo 3D it was an idea as doomed as quad engines.
    Duke Nukem is the same thing, and all voxel engines (which are really the same thing but taken a step further), onyl good for height mapping.
    However, a hybrid engine would eb EXTREMELY useful: terrain normally is perfect for that and a ton of buildings/objecs would work fine for Doom/Duke3D type rendering too. Polygons would thus be used sparingly to supplement that, along with scaled 2D objects, of course -and paper "cut out" textures with 3D perspective. (and that's exactly what Outcast on PC did)

    So a flexible GPU/DSP/blitter set facilitating all of that at good speed, but not nearly as good at pure polygons as the PSX. (the RSP in the N64 might have fir the bill to some extent too, bu I don't think as well -reltively speaking- as the Jaguar's RISC GPU)
    With height mapped spans you have constant Z texture rendering that will never, ever warp, and with voxels you can have very smooth terrain (a voxel engine by extention will allow Doom/Duke3D like spans/stairs/walls/objects I believe as it's a more flexible extension of those methods -rendering vertical columns of pixels) with polygons used sparingly along side all of those.


    That more than implies that there was some sort of resource trade off. Whether it be the "newness" of triangle rendering or not, something compelled very notable engineering firms to go with quads instead of triangles.
    No, just more silicon to design, and "triangles" were not new, they were the decades old polygon standard for 3D.
    6 days older than SEGA Genesis
    -------------
    Quote Originally Posted by evilevoix View Post
    Dude it’s the bios that marries the 16 bit and the 8 bit that makes it 24 bit. If SNK released their double speed bios revision SNK would have had the world’s first 48 bit machine, IDK how you keep ignoring this.
    Quote Originally Posted by evilevoix View Post
    the PCE, that system has no extra silicone for music, how many resources are used to make music and it has less sprites than the MD on screen at once but a larger sprite area?

  12. #12
    Bite my shiny, metal ***! Hero of Algol retrospiel's Avatar
    Join Date
    Mar 2008
    Location
    Cologne, FRG
    Posts
    7,816
    Rep Power
    91

    Default

    Without having actually read the original article, I wholeheartedly agree with Damien McFarren's points as quoted in your review:

    The 32X was "a poisonous tumour that would further erode the company's standing in the global marketplace."

    It "made [Sega] look greedy and dumb to consumers, something that a year earlier I couldn't have imagined people thinking about us. We were the cool kids."

    And yes, the Playstation launch in Japan made "any argument in favor of the 32X ... ridiculous."

    I am 110% sure that working for Sega from 1992 through 1995 was "like watching the Hindenburg in slow motion." With Full Motion Video, Activator and Sega VR I had the exact same impression as a consumer.

    And yes, consumers "voted with their wallets and stayed away."

    - I mean why buy shitty Sega stuff if you could buy Nintendo or a PC ?
    The Mega Drive was far inferior to the NES in terms of diffusion rate and sales in the Japanese market, though there were ardent Sega users. But in the US and Europe, we knew Sega could challenge Nintendo. We aimed at dominating those markets, hiring experienced staff for our overseas department in Japan, and revitalising Sega of America and the ailing Virgin group in Europe.

    Then we set about developing killer games.

    - Hayao Nakayama, Mega Drive Collected Works (p. 17)

  13. #13
    Banned by Administrators 16bitter's Avatar
    Join Date
    Jul 2005
    Posts
    685
    Rep Power
    0

    Default

    Quote Originally Posted by Christuserloeser View Post
    Without having actually read the original article, I wholeheartedly agree with Damien McFarren's points as quoted in your review:

    The 32X was "a poisonous tumour that would further erode the company's standing in the global marketplace."

    It "made [Sega] look greedy and dumb to consumers, something that a year earlier I couldn't have imagined people thinking about us. We were the cool kids."

    And yes, the Playstation launch in Japan made "any argument in favor of the 32X ... ridiculous."

    I am 110% sure that working for Sega from 1992 through 1995 was "like watching the Hindenburg in slow motion." With Full Motion Video, Activator and Sega VR I had the exact same impression as a consumer.

    And yes, consumers "voted with their wallets and stayed away."

    - I mean why buy shitty Sega stuff if you could buy Nintendo or a PC ?
    Thank you. Damn fine job taking Occam's Razor to this interminable fanboy squawk-fest of a thread.

    Some would have us believe that Sega made fine decisions back then. It's just too bad for them that something called 'reality' got in the way.

    Sheath's been plotting his revenge on this esoteric phenomenon, quite obviously, ever since.

  14. #14
    I remain nonsequitur Shining Hero sheath's Avatar
    Join Date
    Jul 2010
    Location
    Texas
    Age
    42
    Posts
    13,313
    Rep Power
    130

    Default

    Quote Originally Posted by Christuserloeser View Post
    Without having actually read the original article, I wholeheartedly agree with Damien McFarren's points as quoted in your review:

    The 32X was "a poisonous tumour that would further erode the company's standing in the global marketplace."

    It "made [Sega] look greedy and dumb to consumers, something that a year earlier I couldn't have imagined people thinking about us. We were the cool kids."

    And yes, the Playstation launch in Japan made "any argument in favor of the 32X ... ridiculous."

    I am 110% sure that working for Sega from 1992 through 1995 was "like watching the Hindenburg in slow motion." With Full Motion Video, Activator and Sega VR I had the exact same impression as a consumer.

    And yes, consumers "voted with their wallets and stayed away."

    - I mean why buy shitty Sega stuff if you could buy Nintendo or a PC ?
    I'm going to need you to put a greater effort into the details of the article before I can reasonably respond. My knee-jerk reaction was to say that you were simply trying to inflame the discussion. I know better than that about you though. So, please elaborate with specifics.

  15. #15
    Hero of Algol kool kitty89's Avatar
    Join Date
    Mar 2009
    Location
    San Jose, CA
    Age
    30
    Posts
    9,724
    Rep Power
    63

    Default

    Continued....

    Quote Originally Posted by sheath View Post
    ROM limitations wouldn't have caused the same thing with "Jupiter"?
    Tons: firstly the ROM bus on the Saturn is FAR faster than the Genesis's 1988 speed, but on top of that you didn't NEED to access ROM nearly as much.
    There's RAM, RAM, RAM, one of the 2 main weaknesses of the 32x, fast, high-bandwidth RAM in considerable quantity, and the biggest advantage of the Jaguar. (albeit the Jupiter as such would be a fair bit more expensive to produce than the Jag, but Sega would be selling at cost and have the advantage of more than 10x the production volume plus already stocking the same chips for the Saturn)
    Thus you could highly compress all data (especially graphics) in ROM and decompress into RAM, depending on what you specifically do the compression could be well over 5:1 (audio would be tougher as ADPCM was the only "standard" scheme in general use unless they invested in a custom codec based on CVSD or a derivative of ADPCM with higher compression ratios). You could go considerably higher with the graphics compression if you opted for lossy schemes rather than strictly lossless stuff. (some Jaguar games employed a custom JPEG format with 14:1 compression)

    And even when you do exceed the available RAM, you have a lot more resource for on-the-fly decompression, so even then it doesn't need to be all decompressed. (and you could also store some moderately compressed data in RAM to facilitate lower overhead but still save space -especially useful for audio, but probably most useful to limit it to samples played on a few channels depending on how many streams the 68k can decompress -if nothign else you could do SFX and maybe some percussion samples that way while the rest of the instruments were decompressed -some of that is up to programming and efficient use of the 68k as well as how computationally intensive the compression technique is)

    I think this understates the similarities in the ways the Genesis-32X and the Saturn's VDPs were designed to render 3D (3D overlayed on 2D). Consider a scenario where the Saturn was going to be in Japan only until late 1995 at the least and the 32X saw full second and third generation software. The ease of down-porting Japanese Saturn software to the 32X and up-porting western developed 32X games to the Saturn would have been greater than adapting either to other platforms. After 1992 we already saw a good number of third parties (especially the crappy ones) developing games concurrently for Genesis, SNES and Sega CD.
    Those are vague superficial similarities: the architectures are very different and the games would have to be significantly modified to work like that. The Genesis and Saturn VDPs are not 2 peas in a pod.
    Actually... the Sega CD's blitter is closer to VSP1 than pretty much any other hardware in the 2 besides the SH2s, and that set-up is a LOT more like the Saturn than the 32x is: the Saturn had VDP1 render to frame buffers which then gets output to one of VDP2's BG planes while the Sega CD renders to frame buffers (word RAM) and updates that to Genesis VDP tiles, albeit that also has hardware sprites to output to as well, not just BG. That's a bit superficial as well, but is surprisingly similar in some other respects. Then Again, the Jaguar is a little like that too: the blitter and GPU render to a frame buffer which the object processor reads to the display, but rather than generating BG planes, the OP generates "sprites" of any size and number. (more or less, not "hard" limit as such, but not texture rendered sprites/objects like the Sega CD, Saturn, or PSX, but an intresting mechanism using a composite of framebuffers manipulated on the screen and output to a single frame along with the blitter's display plane -the blitter than GPU render to OP generated windows from what I understand, though I don't fully understand the nature of the system)

    Could part of the "Mars"-Saturn scenario have been intentionally making it difficult to port games to non-Sega hardware?
    I doubt it, at least in terms of the 32x, and if so it was an idiotic idea as it was more an issue of games getting ported TO the system than the other way around.
    That was certainly something that would turn off 3rd parties and approached Nintendo's anti-competitive policies and restrictions in the late 80s.
    And in the context of the PSX and 3DO it would be a horrible move.
    Now had they made it friendly to program for and port to, but powerful in some key areas that would tend to look like crap on competition: that would be significant. (the one area the Saturn does that is using VDP2's BG capabilities for both detailed 2D BGs as well as the pseudo 3D planes used in many 3D games -like a far more advanced/detailed mode 7 with much more color and up to 3 planes)

    If more is always better, this should not have been a problem.
    How so? It would take a lot of added work to make Saturn quality games from 32x games, though probably fairly reasonable work for minimally enhanced straight ports. (the main enhancements would come naturally from the higher bandwidth and CPU speeds -or lack of using on-the-fly decompression)

    I'm almost certain that Doom's limited PS1 exclusivity was an early example of Sony's heavy-handed anti-competitive philosophy.
    Huh? What about 32x, 3DO, SNES, N64? (albeit the latter was a remake, not a port and the 32x version was 1994 like the Jag so a somewhat different context)
    Sega already had a license for Doom and had published it for the 32x, so that alone should have heavily facilitated it: id ported a lto of games to the Saturn and multi-platform in general, and the Saturn got Quake when the PSX skipped to Quake II. (the Saturn got a re-made custom engine for Quake too as there wasn't enough RAM to do it otherwise, let alone the use of quad rendering vs slower software based triangles)

    Virtua Racing was only a month late to the Saturn's early US launch and is essentially a more advanced version of VR Deluxe. Then there's Galaxy Fight, Ghen War, and Wing Arms or Panzer Dragoon (Star Wars Arcade is a very small step above a rail shooter). Unless the topic of the discussion is the importance of 32X software/brands for the early success of the Saturn, I don't see the benefit of adding these games to the 1995 Saturn library aside from splitting their sales up even more than in reality..
    Except many consider it worse than VR deluxe and wasn't even developed by AM2 like the 32x game.

    Third party support for the Saturn was in fact higher than PS1 support in 1995. For some reason I think more 2D or psuedo 3D platformers in the early Saturn library would have been great though.
    Yes, I didn't claim otherwise, but they needed support months sooner if the May launch was going to work, that was my only point. (not that they should have done the May launch at all, but a best case of making it work if forced)
    The PSX would have been about as bad off if not worse if launched in may.


    All of that is just trying to make the Saturn a mass market product earlier than it was ever going to be. Even the PS1 wasn't mass market that year.
    Yes, but again, that was in the context of Sega having absolutely no choice but the release the Saturn in May and managing the strongest launch and the best PR possible rather than the mess they ended up with.
    Especially in the combined context of the Saturn's Japanese launch date. (again, September would have been fine, and August a good option if they felt jumping Sony by more than a week was necessary)

    You mean tendencies like assembly level software development for multiple CPUs, VDPs and sound processor capabilities and 3D overlayed on 2D?
    Those are all vague generaliztions that the Sega CD would already have done sans leaning SH assembly, the architectures are not even close. And programming in C isn't bad either, there's some critical things that need assembly language optimzations, but in many cases C is OK: of course compiling CPU code and programming to a full system library/OS are 2 very different things. (remember that Wolf32x is in large part programmed in C compiled to SH2 with some SH2 assembly mixed in and 68k assembly as well)

    Getting use to the graphics architecture's specifics would have been critical though (particularly optimizing for bandwidth limitations, trade-offs using advanced VDP2 effects, etc), the CPU architecture is only one facet of the hardware. (not to get into optimizing for the sound system, using software compression and managing with the 68k, working around bugs, and using the geometry DSP)

    Based on Bayless' suspicions I would say that Sega of Japan didn't know whether the Saturn would make the 1994 launch window. Why, or whether, they felt that way could have had to do with any number of organizational, development, contractual, marketing, software or financial reasons.
    Regardless, SoA should hve knows that SoJ was trying as hard as possible to make that window and that there was a strong possibility that the Saturn would indeed be released about the same time as the 32x.

    Or a classic example of human beings misjudging the future and then failing to adequately correct said mistep in the present. There is no indication that SoA was not aware of the Saturn's launch by mid year at least. The facts point to them being nearly unanimously disappointed in that knowledge though. That disappointment may have caused any number of problems regarding PR, software, or even final stage 32X design optimizations.
    They should have been made aware of SoJ's intentions when Mars was brought to them back in January of '94 and managed Mars/32x with that in mind.

    IIRC even Japanese publications early in the year had been promoting plans for the Saturn's release that fall. (Pettus mentions that -actually he mentions even earlier launch dates being promoted, but I don't recall the specific sources... his citation is pretty screwed up in most cases )

    And yet they make the industry go round, right down to what kinds of games are created and how marketing dollars are spent. I feel the need to correct popular misconceptions mostly because I am so bored with the state of the industry today.
    I feel the need because it's ridiculously frustrating not only for average "dumb" gamers/fans to spew BS info, but genuinely interested people gettign misinformed due to the proliferation of that tripe.
    That's part of what makes Kent's book so frustrating: it's not on the worst end, but there's tons of problems across the board: lots of good quotes and interviews, but absolutely horrible fact checking which specifically came up here:
    http://sega-16.com/forum/showthread....307#post303307
    That's Atari-specific, but the problems remain across the board for that book for every single subject addressed. (Curt Vendel and MartyG's knowledge and investigation is just a specific and heavily highlighted example of the problems)

    It's so close to being a really good book and genuine resource had he only done a good deal of cross-referencing and fact checking for it, and of course it's too late to revise it as he no longer owns the rights and is also not interested in the subject anymore.


    I would have clipped this, but I decided to take the opportunity to tell you that I agree with the above completely. I have often been disappointed in "fully 3D" games that might have been done just as well or better with sprite scaler engines.
    I agree too, but it definitely depends on the case: Wing commander is great the way it is, especially for the time, but X-Wing wouldn't have been nearly as good had they opted for sprite/bitmap rendering. (even dropping to the lower detail levels necessary to run on a 386 at the time, it's generally preferable -especially how many ships, especially Ties, cater very well to low polygon count graphics)
    Origin's shift to texture mapped 3D was also timely for 1994, albeit they certainly pushed the envelope for system requirements as typical of them.


    Right, I just haven't seen any discussion on the topic reach a definitive conclusion. Also, the software evidence leans absolutely towards the Genesis being superior in all things but colors and audio. I think this could be attributable to specific capabilities of the Genesis VDP, but it is more commonly attributed to relative CPU prowess in regard to (MD specific) games.
    I also forgot to mention the buss access speeds: the 650x arch can access the main bus in 1 cycle while the 68k takes 4, a substantial difference, albeit the 650x's nature makes it more necessary to access the bus (especially using "zero page" registers -the first 256 bytes of RAM in the address space) and the 68k reads 16-bits at a time vs 8-bits. In particular, the 68k's slow accessing can also be taken advantage of for some neat bus interleaving tricks as the Amiga uses. (allowing 50% bandwidth use by the coprocessors without the CPU taking a hit) A lot of trade-offs so it's hardly simple.
    The SNES's CPU is still a lot slower than it should be, and the RAM at very least should have been at 3.58 MHz rather than forcing it down to 2.68 MHz (so most early SNES games have the CPU running at 2.68 MHz practically all the time vs later games using 3.58 MHz ROM and only RAM being slow), but the CPU easily should have been faster as well given common speed ratings for the 65816 of the time and common DRAM speeds. (it should have been 5.37 MHz at the very least if not 7.16 MHz -clock is a divisible of 21.48 MHz)

    His opinion reflects the common perception in the industry though, and especially late 90s early 2000s conjecture. I doubt Bayless has spent any time arguing with homebrew developers and enthusiasts.
    No, but he could at least have talked to Saturn developers... I've traded a few comments with some former Saturn devs on Atariage. Software rendering was hardly uncommon: Amok ring any bells, or Duke Nukem 3D, or several polygon based games using triangles. (not sure which specific games used software triangles though, but I'd guess several sloppy PSX ports at the very least -like Doom)

    That is a touch sloppy. The SNES could pull from a 32,768 master palette of colors, but could only display eight 15-color palettes simultaneously in the most common modes. I am not aware of the 2048 "Direct Color Mode" ever being used. The SNES could not display its master palette though, and the 32X apparently could, though that ability was equally impractical to the SNES' Direct Color Mode.
    That's what I meant: indexed colors from 15-bit RGB in both cases. In all games (save maybe virtua racing) the 32x used 256 color mode and it didn't have hardware 8-bit texture indexing either. (you'd need to use software to store textures as 256 colors and output them to a highcolor display, which Chilly Willy did for Yetti3D and is what the Saturn, Jaguar, N64, and PSX do in hardware -PSX also supports 4-bit textures and 15-bit textures as does the Saturn, but I don't think the jag does 4-bit as I've seen that on a hardware feature "wish list" before)
    The SNES's mode 3 was the most useful 256 color tile mode and also featured a full 16 color tile layer as well (so very much like the common mode 2, but with one 16 color layer replaced with a 256 color one), but no 2D games I know of actually used it due to VRAM space limits, not to mention bandwidth limit. It's not really all 256 colors as the palette entries are shared among the sprite and 16 color tile layers, so at most you'd hit 241 colors per tile without palette re-loading, but that's pretty damn close. However, for really useful sprite and 16-color tile palettes, you'd have to sacrifice a bit more on that, but probably over 200 colors still. And you could also cut the bitplanes from 8 to 7, 6, or 5 and make it more realistic to use: in those cases, you'd have 128, 64, or 32 colors per tile rather than 256. (in that case, you'd likely specifically organize the palette entries to manage a near full 128/64/32 colors and have the other entries for the 16 color BG and sprites) Dropping to 32 colors per tile would probably have significant disadvantages to plain 16-color layers as you're stuck with those 32 colors, not multiple 32-color palettes (more flexible palette indexing would have been great though).

    Somehow this does not show up in software.
    That's your opinion, but it's a technical fact that the PSX can push more objects than the Saturn due to the high texture mapping bandwidth.
    Again, that would only fully manifest in games with simpler backgrounds as the PSX builds the BG with textures as well vs the Saturn's dedicated BG generator.
    The PSX loses that advantage pretty quickly after a couple full-sized BG layers, let alone trying to simulate other VDP2 effects.
    That's a trade-off of flexibility though, and the PSX was certainly quite capable in 2D. (the N64 might have been more capable, especially with that unified 4MB block, but it was never pushed in that direction and the standard "fast 3D" microcode was hardly optimized for high-speed 2D -the turbo 3D code sounds far more appropriate if still far from ideal)

    That's oversimplified again. The Saturn had a total of 1.5MB of VRAM to the PS1's 1MB. The PS1 had to render everything, including backgrounds and framebuffer, to that 1MB. The Saturn, as you noted, had backgrounds down pat without any hit to that 512 "texture RAM," but it also had 256KB framebuffers for both VDPs. Also, developer interviews and published software show the PS1 hitting a "wall" in 2D animation and color counts and the Saturn not hitting a limit.
    Again, I addressed all of that, and was talking about the heavy advantages of sprite heavy games.
    As I said, take a common 320x224 highcolor game: on the Saturn you have dedicated framebuffers of fixed size, so there goes 512k, the PSX has variable side in shared memory and in this case it's 280 kB to double buffer (if you knew you could manage 50/60 FPS, you could single buffer for 140 kB and not worry about tearing), but normally 744 kB left over for textures. So then you have the dedicated 512 kB for textures in the Saturn, significantly less, but another 512 kB for BG tiles: so if the BG tiles took up more than 232 kB then the PSX would be at a disadvantage, but otherwise have an advantage. (both can use main RAM for added storage to page textures/tiles in to video RAM too though the PSX has 2 MB of 133 MB/s RAM vs the Saturn's 1 MB of 114 MB/s and 1 MB of 57 MB/s RAM, plus the slow RAM in the RAM cart on the Saturn)


    I am sure there are enumerable differences in implementation. I was referring to the overall design idea as opposed to your suggestion of a faster single CPU that likely cost more than double the dual CPU solution. Cramming half again the transistors while doubling the speed is usually much more expensive than two processors assisting one another and *not* fast enough to justify the cost difference.
    It depends on the case for sure: and the SH3 is NOT just a 2x speed CPU, it's a far more powerful CPU in general, somthing getting close to actually manage decent PSX competitive software rendered 3D. (let alone the added flexibility)
    As for cost, more chips on the board with more traces is a significant issue vs fewer chips at higher price, but in the case of a faster SH2, they simply weren't available at the time, so it's moot. The SH3 would have been in the context of a 32x-like concept on an actually competitive level for the 5th gen and still much cheaper/simpler than the Saturn. (it likely could have been cheaper than the N64 too: the R4300 would be fairly cost comparable to the SH3 if not a bit more, but in the context of such a purely software rendering Sega system, the VDP would be very simple and no RSP to worry about -which was about as much silicon as the R4300 plus the integrated I/O, audio, and video generation circuitry)
    But yeah, as I said, dedicated hardware aimed at low cost would be cheaper, with middle ground in using a mix of powerful CPU and somewhat flexible DSP resource to not limit you too much if you didn't want to go Sony's route for 3D. (in hindsight, going with the flow made all the sense in the world, but back in 1993, let alone 1990 like the Jag, that was hardly a clear-cut issue)
    Again, with a DSP and maybe a simpler blitter (along the lines of an enhanced Sega CD ASIC) you'd have lots of flexibility, plus the CPU resource: so you could do triangles, quads, ray casting, various 2D engines, etc at similar performance optimization. (with the blitter you'd have texture mapping at least, for line by line rendering -ray casting can't use that though)

    Well, the entire industry has either missed the topic of dev kits entirely, or fervently believe that Sony revolutionized the industry with them.
    Dev kits had been standard for a LONG time, but the thing that changed was detailed hardware level documentation vs supplying comprehensive high-level libraries.
    However, PC development was already shifting that way: yes indeed you had such dev tools on the PC in the context of full APIs and OS level programming mainly in the context of windows games emerging, especially after Windows 95 came out. So it paralleled the PSX fairly closely, but was independent.

    But again, in terms of high-level programming in general (especially C), regardless of APIs and such, C programming on an application level was getting rather common by 1993/94 and C along with other high level languages had already been common among a broad array of applications for computers for over a decade. (the games/demo market was pretty much the last to really hang on to by hand assembly)
    Atari/Flare didn't have the advantage of designing the jag late enough to realize that as it was started in '89 and the core logic was set in '90/91 with the silicon design and debugging following that and most being completed by the end of 1992 with some ongoing tweaks in early '93 prior to freezing it for production. (I might be mistaken on that and it might have been set by late 1992 and I recall some mention of dev systems being released by late '92... though it's not uncommon to do tweaks after releasing the dev kits -that happened with the 32x having buggy dev units and clean consumer units -hence the lack of DMA audio)

    Yes, and Sega themselves teamed up with Microsoft for Direct X during the Saturn's days for Sega PC softs, and then again with WinCE on Dreamcast. I'm not arguing the trend itself, or even its merits (lower cost development was a plus for everybody involved).
    Sega wasn't only using direcX either, but several other early APIs: the software renderers tended to be directX though.
    I need to check again, but I'm almost positive that most/all Saturn ports supported ATi's RAGE line as well, with Direct3D being pretty weak at the time and Glide not even being out until '96 (plus you had the weak S3 ViRGE used by Diamond's Stealth 3D). Supposedly they also supported the NV-1 (Diamond's EDGE 3D which also had onboard audio and saturn controller ports) with quad rendering, but that card was never popular. (ViRGE was onyl popular due to the low cost, but had crap 3D, but Rage was good and probably the earliest real competition for 3DFX's Glide based cards)
    We had a Rage pro card in the late 90s and it ran BUG! especially well.

    Since most of that will never be known, release dates are the only real indicator. The Saturn, with no high level dev kits, managed exclusive versions of almost all of the popular PS1 titles within a month or two (i.e. they weren't a reaction).
    Yes, so obviously the Saturn had support early on, I never argued that, and that's also a big reason that made the May launch even stupider. (and may have actually weakened the lineup for September by rushing some developers to release sooner than planned)

    This I would totally agree with. There were too many rumors in 1991 of a System 32 based console for it to be complete magazine fabrication. I suspect that, like the 32X and "Jupiter", the Sega CD's popular reception curbed that project. Alternately, Sega might have simplified the 2D aspects and started focusing on how to create a 3D home console capable of porting their arcade games.
    Maybe not simplified the 2D aspects, but probably just stopped focusing on them if anything and worked on cost reducing/consolidating the hardware as much as possible.
    Given when the SH1 and SH2 were even released, the SH1 couldn't have been added until 1992 (the first silicon wasn't even made until September of '92) and the SH2 couldn't have been adopted until late 1993 at the earliest. (first silicon wasn't tested until October 1993), so even pre-production versions wouldn't have been available before then. (the SH3 didn't come into play until 1994)
    http://www.hotchips.org/archives/hc6...S4/HC6.4.2.pdf


    That is interesting. Integration to a single chip early on usually implies some sort of shared responsibility. I had not noticed a second digital signal processor in the Saturn previously mentioned though. So the Saturn had the SH-1, Motorola 68EC000, and a DSP apart from the dual SH-2s, Dual VDPs, SCU and geometry DSP. Not to mention a much higher product quality CD-ROM drive than the PS1's and more memory overall. Hot dang they went all out, I wonder if we will ever see notes on why every component was chosen.
    No, it doesn't imply that at all, it just means they were trying to integrate as much as possible... The NES had the I/O and sound hardware integrated into the CPU from the start, the SMS VDP had the PSG integrated from the start, the Atari 2600's TIA had the audio, joystick trigger inputs, and analog ports on it on top of video, POKEY had all the keyboard I/O plus audio, the Sega CD's ASIC had the blitter, DRAM control/interface logic, an Genesis interface logic onboard, the Amiga's sound chip also had the floppy disk controller built-in, etc, etc, it's just cost saving where feasible. Obviously you'd preferably add it to an existing custom chip on the board and if the VDPs were already big and pushing the practical limits fo the silicon they were on, another chip would be used.
    Sometimes it's related chips, sometimes totally unrelated.

    Right, CD-ROM games with less RAM means shorter levels or more frequent loads. Thus to allow levels of comparable lengths to contemporary cartridge games the Sega CD needed more than just CD cache.
    Yes, so load times are a totally independent factor.
    Without that cache, the load times would be much longer: ie 2x as long (or more) than the Super CD.


    That was primarily where I was coming from. This is much the same as the SNES versus Genesis in-game performance discussion. We do not have pertinent facts regarding what percentage of the SNES CPU was being used in any production level software to say that it "could have done more."

    What would we have sacrificed in regard to what the Genesis CPU's in game responsibilities to manage loading efficiently? Just more time with a blank screen?
    Nothing would be sacrificed in terms of plain loading, but the tough thing would be games doing on-the-fly loading: ie loading while running the game, especially streaming/decoding video and and (no CD-DA) audio. CD-DA is handled by the CD chipset itself, the rest is managed by the CPU.
    Again, a cheap MCU should have avoided that completely with the only disadvantage being loss of a general purpose coprocessor on the level of the 2nd 68k when not loading data. (the same as would have been true for the Saturn's SH1 had it been allowed to be used for other purposes when not managing the CD)

    As for the SNES, pretty much everything on the Genesis could have been done on the SNES, it's just a matter of optimization and trade-offs in general. Later SNES games definitely had the advantage of a 33% faster CPU for most of the time. (ie faster ROM as most code is run straight from ROM and the RAM is as slow as early ROM while later ROM pushed the full 3.58 MHz) The exception of non-enhanced games on the SNES would be the 3D games running far worse, or at least the polygonal ones. (the cheap DSP-1 should have helped a ton though, not Super FX, but far simpler/cheaper to add, and available from 1990 onward, in fact with the DSP1 alone the SNES might be better at polygonal 3D) For ray-casting and scaling, that's another matter too, but Wolf3D is pretty damn impressive. (a nice trick upscaling mode 7 like that too)

    There they are again, packed pixels versus planar graphics. We're talking about tiles versus full screen frame updates? You are correct that the NEC compression comments emphasized more data per disk and not more data per level. But the context was "better games."
    No, packed pixels vs planar are 2 fundamentally different methods of storing pixel data: packed using linear pixel storage: ie for 4 bit pixels you have 4 bits for one followed but 4 bits for the next, but for planar graphics you have bitplanes, basically a group of 1bpp bitmaps that get composited: fo for 4bpp you have 4 bitplanes, 4 separate 1bpp regions in memory that get composited.
    We discussed this before:
    http://sega-16.com/forum/showthread.php?t=9265

    A framebuffer vs tilemap display is a totally separate issue: ie the Amiga and SNES both use bitmplanes (and NES, PCE, SMS, EGA, ST) while the Genesis and 32x both use packed pixel graphics (as does CGA, Atari 8-bit/5200, VGA, and pretty much all video hardware from the mid 90s onward).


    It would be much more conducive to conversation if you would summarize the points and cite any sources used. As it stands, the facts I have seen indicate that Sega's games were generally commercial failures with few that drew consumer interest like NES franchises could. VGChartz (I know, I know) had an article a couple of years ago that claimed to have found total software published for each system, and Genesis software was still way behind. That article has long since vanished though.
    I pointed to that discussion so you could see the context and contend with it there... there are several specific articles listed pointed to a 55% market share in 1994, but again see the other thread for details or for additional comments.

    I hate to hurt any feelings, but Europe's market size was relatively insignificant. Except in the sense that European developers focused on locally popular platforms, software or hardware sales there was barely a dent in the Jp/US market.
    Europe made up over 25% of Genesis/MD hardware sales... (if you go by the ~35 million worldwide figure, though it might have been more than that, especially with Japanese WonderMega sales among other things -there's mixed evidence that the Womdermega may have been more popular than the standalone genesis -namely that Sega CD software made up the vast majority of Japanese 16-bit Sega sales at the time, and then the late 90s tectoy sales...)


    Certainly, but only at a retail price similar to any individual cartridge game. This does not bode well for a revenue problem like Sega was facing for the previous four years.
    And how would that differ at all with the 32x games at identical prices/cost? (Sega CD games would obviously be lower prices, so it would be a matter of volume to increase revenue)
    That would have been one thing making compilations unattractive: reduction in revenue compared to selling individual games. (unless the games stopped selling otherwise and a re-release wouldn't help... low-cost non-compiled CD releases would be more attractive for Sega for that reason)

    SoA could have prepared a much more concise marketing stance regarding the cost effectiveness of the 32X for existing Sega owners. As it was they more than implied that Saturn was out of reach both in price and in release date until the Saturn's Japanese launch. After that point the entire mood regarding Sega in general turned for the worse in the media especially.
    Yes, and again, Sony's accelerated price reduction severely reduced the impact the 32x's cost would have, and the Jaguar's. (and 3DO's) It seems quite possible that if Sega kept their normal tends of selling at or near cost, the Saturn probably wouldn't have even dropped to $300 until mid 1996, not $200 until '97, etc.

    It probably was Nakayama's fault alone. Even this article alone describes him going against SoJ's "consensus" for the Saturn for the 32X's creation. Whatever he had in mind though, *something* caused Sega to launch the Saturn too early and that thing was not the PS1 alone.
    I got the impression that SoJ had wanted a successor out even before 1994, but it delays forced it to be delayed.

    PR across the industry continually lost credibility from 1994 on. I don't believe in coincidences. Sega merely contradicted themselves with the 32X and Saturn, other companies were actively involved in public deception.
    The 32x and CD were contradictory as well, but the Saturn made it worse. It became a lot easier to criticize after the fact, but it really was like that to a degree.
    At least in the context of the SVP it would be 1/3-1/2 the price of the 32x alone, or integrated into carts. (and maybe even get integrated into a duo system)

    That is, in fact, the only reason I think Jupiter might have worked somewhat. But for me personally, knowing that I was buying the lesser hardware, I just would have saved up for the full version unless there was an incremental option of equal cost. So now we're discussing a Jupiter, Jupiter CD-ROM, and a Saturn.
    I still fundamentally disagree about the nature of CD game development... however I will concede that the high-price nature of CD based systems demanded premium software, but I think low-cost software would be a huge attraction: just like buying a computer for 4-5x that of a game system and saving a ton in the long run due to the low price of games.
    But especially, a modestly priced system with CD-ROM would have made the ideal budget system: low cost media, low risk development due to that media, high capacity, multimedia capabilities, etc, etc. That's what the Sega CD/Duo might have been if they could push that angle.

    Zyrinx was only one Scavenger team, Lemon's demo at the end of the video is almost definitely showing part of AMOK's engine. It is also worthy of note that even within the relatively low quality video the framerate dips during the Zyrinx portion during the texture mapped gouraud section. Similarly, Among the Zyrinx demos there is very little panning or "strafing", so this might be indicative of framerate limitations.
    Maybe Amok, maybe not, all voxel engines are very similar in nature, so unless you specifically recognize the level design, it may not have been Amok. (and even using the same engine, it could be a different game)
    It's a bigger shame that the Jaguar hadn't had that sort of rendering pushed like crazy from the start. (off topic, I know, but that always pops to mind, especially after seeing Phase Zero in action -albeit that's well beyond what earlier, less optimized engines would look like, but Amok level voxels should have been reasonable)

    Sega just never marketed, and probably never envisioned, the Sega CD for all of these purposes. You have seen how particular actors get type cast for a singular role when they are capable of much more. Similarly, if the Unreal 3 engine had been demoed with a 2.5D platformer instead of an FPS and Gears of War it might not have ever been used for such diverse functions.
    There's absolutely no reason they shouldn't have by 1993. 3rd parties were doing just that to a limited extent: texture mapped polygons used sparingly.
    It's ridiculous that Wolfenstein 3D didn't get ported among a vast number of other things.

    I doubt that Sega failed to understand the Sega CD's full capabilities. So, I can only suppose that the Genesis radically superior hardware and software sales made the difference if the expense of 3D development did not.
    Then I guess they were crazy by not using the CD to trump Super FX and the SVP+CD to approximate the 32x (with trade-offs) without spliting the market...
    Especially not shifting that in 1994 when the CD could have been aimed at as an emerging affordable and eventually nearly ideal budget system. (especially a $200 duo in 1994... or maybe a little mode if they integrated the SVP mapped compatibly to a corresponding SVP add-on)
    Again that would even impact FMV quality as the SVP should have aided with decompression beyond what the Genesis/CD CPUs couls handle.
    Last edited by kool kitty89; 10-09-2010 at 01:22 AM.
    6 days older than SEGA Genesis
    -------------
    Quote Originally Posted by evilevoix View Post
    Dude it’s the bios that marries the 16 bit and the 8 bit that makes it 24 bit. If SNK released their double speed bios revision SNK would have had the world’s first 48 bit machine, IDK how you keep ignoring this.
    Quote Originally Posted by evilevoix View Post
    the PCE, that system has no extra silicone for music, how many resources are used to make music and it has less sprites than the MD on screen at once but a larger sprite area?

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •