Review of Retro Gamer's 32X "Retroinspection"

[Chilly Willy]

There is a setting in the 32X that determines which layer is in front, the 32X or the MD; after that, you can flip that on a per pixel basis with the priority bit for the color on 32X pixels.

Is per-pixel priority only supported in 15-bit mode, or can it be used in 8-bit mode too? (and wouldn't that sacrifice 1/2 the palette?)

For 256 color mode, it's the msb of the palette entry, so it can be any or all of the 256 colors.

[kool kitty89]

For 256 color mode, it's the msb of the palette entry, so it can be any or all of the 256 colors.

Hmm, so that would allow you to effectively have 2 128 color planes, right? (1 low priority, 1 high priority) That's actually something the SNES's mode 7 supports. (not sure if any games take advantage of that, but it could be useful for some things)

[Chilly Willy]

Hmm, so that would allow you to effectively have 2 128 color planes, right? (1 low priority, 1 high priority) That's actually something the SNES's mode 7 supports. (not sure if any games take advantage of that, but it could be useful for some things)

No, it's in the PALETTE ENTRY, not the palette index (pixel). The palette entries have the exact same format as 15-bit mode - one priority bit, five blue bits, five green bits, and five red bits.

[kool kitty89]

No, it's in the PALETTE ENTRY, not the palette index (pixel). The palette entries have the exact same format as 15-bit mode - one priority bit, five blue bits, five green bits, and five red bits.

But you could set up the palette index such that the msb or lsb could effectively be used as a priority bit. (as well as more specific high/low priority sorting on a per color index basis)

[Chilly Willy]

But you could set up the palette index such that the msb or lsb could effectively be used as a priority bit. (as well as more specific high/low priority sorting on a per color index basis)

Oh, I see what you mean. Yes, you could set it so that the first 128 entries in the color ram have the priority bit clear and the second 128 entries have it set. Or any combination thereof. It's more a per-color priority than per pixel when you think about 256 color mode that way. Any, all, or none of the 256 colors can be set as priority.

[sheath]

There are some interesting tidbits in Melf's interview of Scot Bayless, in relation to Retrogamer's interview and the current understanding of the 32X's development. It is important to note that Bayless has been consistent with his account across both interviews, something in which Kalinske is somewhat the polar opposite. What I thought interesting in this new interview is that Bayless describes his job experience as a massive rush to getting the Sega CD up and running that was effectively cut off by the 32X's break neck development cycle. Perhaps more importantly though, is Bayless' insinuation that the final 32X was dramatically stripped down and cost cut from the original design.

Bayless doesn't give any specifics in this interview, but mentions that Sato, and Sega of Japan, forced SoA to keep the 32X below a certain cost, and that there was an original more powerful design by Marty Franz. The final 32X being inferior to the original design is attributed as the ultimate reason for the add-on's failure. I fail to see how a more expensive 32X would have succeeded even if it did made the Genesis as powerful as the Saturn. If anything, the high price of the real world 32X in the UK and South America doomed it to not only failure, but everliving ire from Genesis owners to this day. It is almost impossible to convince people from these regions that the 32X was easily cost effective in the US, in relation to the Jaguar and chipped SNES games especially.

[Chilly Willy]

I don't doubt that the 32X was stripped to be as cheap as possible. I'd bet the SDRAM was probably 32 bits wide instead of 16, but they cut the memory in half; then all the ads and boxes all say the rotation and scaling is in HARDWARE... which got cut as well.

Oh, and your link is WAY wrong!

[sheath]

Hahah, oops. Fixed it.

That is an interesting point about the hardware scaling and rotation. I always just assumed that was marketing in direct competition to Nintendo's Mode 7 nonsense. I didn't even know what a frame buffer was back then, but I wanted the Genesis to have "scaling and rotation" that improved with each upgrade.

Would the 32X SH-2s have seen significant performance increases with a 32-bit RAM bus. I'm supposing they do all of their calculations at 32-bit even with 16-bit RAM. I would think that the 32X would need more RAM, or faster ROM, before it would need a 32-bit wide bus.

[kool kitty89]

Would the 32X SH-2s have seen significant performance increases with a 32-bit RAM bus. I'm supposing they do all of their calculations at 32-bit even with 16-bit RAM. I would think that the 32X would need more RAM, or faster ROM, before it would need a 32-bit wide bus.

Not everything would be done with 32-bit computations (a good assembly programmer -or compiler- would use the fastest operations practical for a given task -so if 16-bit operations were faster and reasonable, then those would be preferable), but yes the 16-bit bus width has nothing to do with that whatsoever. (in fact, due to the burst-mode access limitation, SDRAM is accesses in bursts of 16 bytes iirc -length of a single cache line- so you waste a lot of time manipulating smaller chunks of data in SDRAM as you always need to access the entire burst/block regardless of whether you only need a small chunk of that -or single 32 or 16-bit word; iirc, these bursts take 12 SH2 cycles to complete where a single, random 16-bit read/write should take around 3 or 4 cycles had they been supported -and a 32-bit read/write on a 16-bit bus should have taken one additional cycle -sequential page-mode accesses could be even faster at 1 cycle per 16-bit word accessed within the same row in RAM, which would be useful for things like texture fetching and such, where you'd read data from a limited region in RAM and write to the framebuffer)

Assuming they used the same restricting burst-access scheme for SDRAM, the advantage of 32-bit width would be double the bandwidth and having less overhead for bust accesses as well. (it should take 4 cycles less per burst to read the same 16 bytes since you'd only have to read 4 32-bit words rather than 8 16-bit words)
However, the framebuffers would still only be 16-bits wide (and rather slow), so that would remain a major bottleneck too.

I don't doubt that the 32X was stripped to be as cheap as possible. I'd bet the SDRAM was probably 32 bits wide instead of 16, but they cut the memory in half; then all the ads and boxes all say the rotation and scaling is in HARDWARE... which got cut as well.

Stripping it to bare minimum cost is one thing, but actual cost to performance ratio is another thing entirely. (ie you can strip down a given architecture to the cheapest practical configuration, but that may still be far less efficient than another design of similar manufacturing cost but overall superior performance) Aside from just "better" engineering (or catering to the ideal feature set for the time), there's the more universal issue of R&D investment.

Making heavier investments in the actual design/engineering work can make a massive difference in overall quality and performance within given cost constraints, albeit such investment isn't just money and manpower either, but also time. Time is one thing the 32x was definitely not given for its design. (approximately 6 months from napkin schematic to preproduction dev units, that's even more extreme than things like the Atari ST)
One could argue that they could have cut corners and accelerated development by making use of existing designs (building on parts of the Saturn and/or MCD, etc), but that's practically limited. (especially since few of those really catered to a higher performance system that would have cost similarly to the 32x)

Ideally, a 32x-like add-on for the Genesis (ignoring the practical disadvantages of marketing such an add-on relative to the Saturn, etc) would probably have been something architecturally closer to the Jaguar in as far as using a wide unified bus with cheap commodity RAM and a single relatively dense custom ASIC (somewhat like TOM in the Jaguar) containing all the interface and RAM control logic, embedded CPU/GPU, blitter, sound and I/O hardware. (so more like the hypothetical streamlined Jaguar Crazyace and I have been discussing in the 5th gen tech thread -in terms of omitting JERRY and cutting out the Object processor line buffers and CRAM from TOM in favor of a simplified system using a single ASIC with blitter and GPU/CPU with audio DACs and I/O logic added to TOM)
Of course, the Jaguar had a lot more time and R&D investment made in it than the 32x (albeit also with the disadvantage of having the architecture laid out in 1990, with limited foresight), so the context is different there for sure.

And as I've said before, in terms of getting a design from napkin sketch to production-ready in 1/2 a year, there's probably not a whole lot more that could have been done, especially with the constraints of it being an add-on. (albeit, from that standpoint, there's a bigger argument for not designing such an add-on at all, and that any such add-on would be much more practical to limit to much lower cost/complexity -like the SVP- and even then be arguably impractical for mainstream market success -especially on top of the MCD and with a next-gen system on the horizon)
Perhaps it wouldn't have been too unreasonable to use plain FPM DRAM for main RAM rather than SDRAM and make it 32-bits wide (probably still cheaper -or allow more like 1 MB of RAM), though that would increase the complexity of the DRAM controller/interface if they wanted to keep speed up (rather than limit it to framebuffer DRAM access speeds). But a decent FPM DRAM controller could actually end up giving better performance than the SDRAM allowed by not limiting the SH2s to bust accesses. (so single word reads/writes would be much faster)

Besides that, having a single 512kB 16-bit wide SDRAM chip (like those used in the Saturn) would have been more important than having 256 kB at 32-bits wide . . . especially if they didn't restrict it to burst reads/writes. (so single word accesses would be much faster -with the speed of the SDRAM being used, probably just 3 or 4 SH2 clocks per read/write)

And in terms of raw cost in the 32x's design, the 2nd SH2 was probably a major waste compared to investing in a larger custom ASIC with hardware graphics acceleration capabilities. (again, R&D could have factored in there for sure, but if manufacturing cost was the only factor, then more custom silicon dedicated to hardware acceleration -blitting/affine rendering/RGB blending/shading/etc, let alone polygon warping or rasterization- would be much more useful than a 2nd SH2 for 2D or 3D games; even a direct derivative of VDP1 might have made sense -using the framebuffer banks similarly to the Saturn and contending with the SH2 for textures in SDRAM or adding a dedicated texture RAM bank as well)


Getting a bit further off on this tangent, there's the argument that, if a lower-cost/interim/early next-gen console was truly felt to be needed (beyond the MD/MCD or SVP-class enhancements), then a standalone system rather than an add-on could have made much more sense. The 32x itself wasn't too far from being a standalone console (just a bit of I/O hardware and actually reduction in complexity by omitting the MD interface and Genlock hardware), and removing the add-on requirement should have allowed for more emphasis on other features (even with the limited R&D timeframe), though anything very close to the 32x would still have been pretty weak as a real 32-bit gen competitor and it would still have just been one more incompatible platform on the market. (which was one of Sega's bigger problems in that period)
So, as I've also argued before, in the context of an early/lower-cost next-gen console instead of the 32x (with all else the same in 1994), the most sensible option would have been something like the so-called Jupiter concept. (a stripped down, fully forward-compatible Saturn derivative with less RAM and no CD drive/interface but retaining provisions for expansion to Saturn spec as well as using carts that would directly play on the full Saturn as well -so still splitting the market somewhat, but allowing a lower-end option accessible to many more users in the 1994~96 timeframe and retaining much more cross-compatibility for next-gen hardware as well as software development -very straightforward ports to Saturn/Jupiter)




Beyond that, you get into the bigger arguments of Sega's management, the design of the Saturn itself, arguable nature/practicality of the Sega CD's market presence (and associated marketing/management), and various other historical business/marketing/management related issues. (along with technical issues)

[Benjamin]

So, as I've also argued before, in the context of an early/lower-cost next-gen console instead of the 32x (with all else the same in 1994), the most sensible option would have been something like the so-called Jupiter concept. (a stripped down, fully forward-compatible Saturn derivative with less RAM and no CD drive/interface but retaining provisions for expansion to Saturn spec as well as using carts that would directly play on the full Saturn as well -so still splitting the market somewhat, but allowing a lower-end option accessible to many more users in the 1994~96 timeframe and retaining much more cross-compatibility for next-gen hardware as well as software development -very straightforward ports to Saturn/Jupiter)

I think you're far too casually brushing off just how terribly it'd split the market, but you are trying to make the best of an impossibly bad situation. Fact remains that the 32X should never have come to market. Poor planning by Sega -- that the company was surprised that Saturn was ready in '94 is insane -- is the culprit here. You can't fault the technology. There's a definite argument to push for an all-in-one add-on to avoid the high SVP pricing seen in Virtua Racing, but that add-on cannot be $150. You lose the argument that it benefits consumers doing that, and then you fail to make significant penetration to get developers to sign on to support it.

Jupiter would have split the market such that the true potential of the Saturn would never be realized as developers target the LCD to ensure everyone can buy their games. The Saturn was already outclassed by early PSX games visually; I'd hate to think what comparisons to 32X-ish titles would have done for sales and the brand. No one would buy Jupiter. There would be no incentive.

[TVC 15]

Bayless doesn't give any specifics in this interview, but mentions that Sato, and Sega of Japan, forced SoA to keep the 32X below a certain cost, and that there was an original more powerful design by Marty Franz. The final 32X being inferior to the original design is attributed as the ultimate reason for the add-on's failure. I fail to see how a more expensive 32X would have succeeded even if it did made the Genesis as powerful as the Saturn. If anything, the high price of the real world 32X in the UK and South America doomed it to not only failure, but everliving ire from Genesis owners to this day. It is almost impossible to convince people from these regions that the 32X was easily cost effective in the US, in relation to the Jaguar and chipped SNES games especially.

Don't mean to nitpick but I thought the interview made clear that it was SOJ top brass and not Sato's intention himself to keep costs down, Sato actually by all accounts thought the original 32X was a neat little design.

Bayless:

I wasn’t privy to the private conversations inside SOJ about how that went but I’d be willing to bet a buck that Sato was just as disappointed as anyone about that.

Still the next big topic to discuss for another few years before we get such a substantial interview like this one again will be how the 32X was scaled down from its original design and then have Kool_Kitty89 hypothesise what it could have been like in some meaty posts.

Hey, I also wonder if we'll get another writer using a few interview sources, and some old magazine clippings to hypothesis how the 32X was originally designed, and then afterward have there subjective narrative history qouted as the verbatim truth! *Cough Pettius* *Cough Saturn*

[sheath]

I think you're far too casually brushing off just how terribly it'd split the market, but you are trying to make the best of an impossibly bad situation. Fact remains that the 32X should never have come to market. Poor planning by Sega -- that the company was surprised that Saturn was ready in '94 is insane -- is the culprit here. You can't fault the technology. There's a definite argument to push for an all-in-one add-on to avoid the high SVP pricing seen in Virtua Racing, but that add-on cannot be $150. You lose the argument that it benefits consumers doing that, and then you fail to make significant penetration to get developers to sign on to support it.

This is just way too broad of a statement to be accurate. For one, $150 is $15 per ten games, or $7.50 per twenty games, it's hardly a horrible upgrade price and I would dare to say that people pay more to get less while having to completely dump their old hardware, and software, in the handheld market. The 32X was badly managed yes, but claiming it should never have existed because it was badly managed requires a lot more than assertive statements.

Jupiter would have split the market such that the true potential of the Saturn would never be realized as developers target the LCD to ensure everyone can buy their games. The Saturn was already outclassed by early PSX games visually; I'd hate to think what comparisons to 32X-ish titles would have done for sales and the brand. No one would buy Jupiter. There would be no incentive.

Splitting the market isn't necessarily a bad thing. If customers want a budget system with entry level 3D and the premiere 3D consoles were going to take three to four years to reach the mass market, then keeping the Genesis as the budget console, while giving that already existing userbase the option of 3D gaming with an add-on, makes sense.

I think a cartridge based Saturn would not have worked at all though, my pocket history rewriter on my new dual core HTC smart phone says that the 32X plus Neptune could have generated more revenue for Sega in the critical years of 1994-1996.

Don't mean to nitpick but I thought the interview made clear that it was SOJ top brass and not Sato's intention himself to keep costs down, Sato actually by all accounts thought the original 32X was a neat little design.

Bayless:

I wasn’t privy to the private conversations inside SOJ about how that went but I’d be willing to bet a buck that Sato was just as disappointed as anyone about that.

I suppose my wording was a little vague, it sounded like Sato was in favor of the 32X design, the original one, especially since he gave up his design for it. It also sounded like Sato was the primary liaison with SoJ on the 32Xs development and as such was the one who handed over the decision from SoJ management about stripping the 32X down.

Still the next big topic to discuss for another few years before we get such a substantial interview like this one again will be how the 32X was scaled down from its original design and then have Kool_Kitty89 hypothesise what it could have been like in some meaty posts.

Hey, I also wonder if we'll get another writer using a few interview sources, and some old magazine clippings to hypothesis how the 32X was originally designed, and then afterward have there subjective narrative history qouted as the verbatim truth! *Cough Pettius* *Cough Saturn*

If I fill pages upon pages with direct magazine quotes will that count?

[Da_Shocker]

Not this shit again lol.

[Benjamin]

This is just way too broad of a statement to be accurate. For one, $150 is $15 per ten games, or $7.50 per twenty games, it's hardly a horrible upgrade price and I would dare to say that people pay more to get less while having to completely dump their old hardware, and software, in the handheld market. The 32X was badly managed yes, but claiming it should never have existed because it was badly managed requires a lot more than assertive statements.

It should not have existed since Sega already had plans for the future in the Saturn. This is more than "assertive statements" but rather based on known history, the results culled from its launch, and its subsequent failure as noted in this most recent interview at that. This isn't even hindsight as there was much press at the time of its launch speculating that it was little more than a stopgap maneuver and, again noted in the interview, much debate internally at Sega as to the 32X's worth with the Saturn launch around the corner.

The price of a standalone hardware upgrade drops as you purchase an entire library, but that goes for anything. How many consumers would you think would have interest in purchasing an entire add-on's library? How many people bought every since Sega CD game back in the day? I find that a silly point to make. At no point is it feasible to consider consumers to react in such a manner. People make cost-benefit decisions on their purchases, and the 32X library is (I believe) safely out of the range for what most people would consider making the add-on worth buying and far less so at launch.

Splitting the market isn't necessarily a bad thing. If customers want a budget system with entry level 3D and the premiere 3D consoles were going to take three to four years to reach the mass market, then keeping the Genesis as the budget console, while giving that already existing userbase the option of 3D gaming with an add-on, makes sense.

I think a cartridge based Saturn would not have worked at all though, my pocket history rewriter on my new dual core HTC smart phone says that the 32X plus Neptune could have generated more revenue for Sega in the critical years of 1994-1996.

There is nothing in history which supports that a company can successfully juggle multiple platforms in such a manner. The problem here is that consoles are not TVs, and marketing them to different markets ignores the fact that content generation is going to be limited, and the company will have to pick and choose which platforms to support. There's no way all three platforms retain solid support, and by splitting up your base you leave it less desirable for third parties, too. It's bad enough now with various companies issuing their own platforms, but then to have three at once competing with each other from a single company? If I were a developer, I'd jump ship in a heartbeat. Too much confusion. Too many programming and licensing headaches. Too few consumers due to the split market, even among Sega fans.

[sheath]

It should not have existed since Sega already had plans for the future in the Saturn. This is more than "assertive statements" but rather based on known history, the results culled from its launch, and its subsequent failure as noted in this most recent interview at that. This isn't even hindsight as there was much press at the time of its launch speculating that it was little more than a stopgap maneuver and, again noted in the interview, much debate internally at Sega as to the 32X's worth with the Saturn launch around the corner.

Common journalistic history definitely makes this assertion, that does not mean it is historical fact. It means that journalists have propagated the idea the way they do all myths, for the express purpose of making the other major platform owners look "smart" and "better". It is also historical fact that Bayless, in his previous interview with Retrogamer, explained the 32X's origins as a "stop gap" 'tween add-on for the existing Genesis market that was intended to keep that category of Sega's revenue up for two to three years while the Saturn took the same amount of time to reach the mass market. If Sega had stuck to that plan the 32X could have been as successful as any "good" add-on ever was, but Sega panicked, jumped the gun with the Saturn and confused their own marketing teams about what the 32X actually was, which was a Genesis add-on.

If Sega had stuck to the plan it simply could have rode the residual sales of the Genesis and Sega CD and 32X and their software while the Saturn took its time ramping up. It isn't just the TV industry that maintains different products simultaneously, all business diversifies or dies.

The price of a standalone hardware upgrade drops as you purchase an entire library, but that goes for anything. How many consumers would you think would have interest in purchasing an entire add-on's library? How many people bought every since Sega CD game back in the day? I find that a silly point to make. At no point is it feasible to consider consumers to react in such a manner. People make cost-benefit decisions on their purchases, and the 32X library is (I believe) safely out of the range for what most people would consider making the add-on worth buying and far less so at launch.

I wasn't talking about the entire library, I was talking about ten to twenty games justifying the cost of the hardware. Meanwhile the $300 PS1 and Saturn, released the next year, needed far more games to reach a similar price per game value. Thus the 32X had a relevant value proposition at the time it was released and in 1995 when it invariably would have been even cheaper.

There is nothing in history which supports that a company can successfully juggle multiple platforms in such a manner. The problem here is that consoles are not TVs, and marketing them to different markets ignores the fact that content generation is going to be limited, and the company will have to pick and choose which platforms to support. There's no way all three platforms retain solid support, and by splitting up your base you leave it less desirable for third parties, too. It's bad enough now with various companies issuing their own platforms, but then to have three at once competing with each other from a single company? If I were a developer, I'd jump ship in a heartbeat. Too much confusion. Too many programming and licensing headaches. Too few consumers due to the split market, even among Sega fans.

Sega did it previously with the Master System, Game Gear, Genesis and Sega CD. Sony did it with the PS1 and PS2, and then again with the PS2 and PS3 and PSP plus the PSPgo and now Vita, Nintendo did it with the Gameboy, GBC, GBA, NES, SNES, N64, DS and Gamecube, Wii and DSlite, and now the Wii and 3DS and DS, Microsoft is doing it with the 360, Windows 7, Games for Windows Live, Kinect, Kinect for Windows, Xbox Live, XBLA, and XNA. Journalistic history is putting the blinders on to these facts, as with all marketing the most often repeated views tend to stick in peoples' minds as "true".