Comparison of 6th generation game console hardware

[stu]

Notice the caveats, "exaggerated" and "an urban myth" based on Kuturagi's claims (eg. individual grains of wood being rendered in a door, etc). The PS2's polygon counts are taken from developer quotes and Sony themselves, do you contest those sources? Now where do you see me claiming that the Dreamcast was more proficient at pumping out polygons, in game or in theory? By citing actual game polygon figures? Obviously every Dreamcast game did not hit these peaks, the games themselves did not average those figures either..

You know, you aren't posting on some hardcore Playstation site where they they mindlessly believe EVERYTHING Sony says, I already know that most of Sony's Hype was bullshit and I wasn't specifically referring to what Kutaragi claimed. I was more referring to where you said "Why? The entire generation was floating around 1 Million polygons per second, especially outside of Racers and FPS." you were quite clearly indicating that you believed that since the the entire generation never got above 1 million polys/sec that the Dreamcast could quite easily keep up with the PS2, when it has already been proven that this 1 million polys/sec max is untrue, yet instead of accepting this and moving on, you attack the people who are refuting your claims by accusing them of "flaming and trolling" you. How about this quote from you? "The bottom line is the Dreamcast was more capable of surpassing 3 million polygons per second than the PS2 and the entire generation floated around 1 million with a few exceptions on each platform." That is plainly incorrect, are you now stating you were wrong in that claim?

Okay, let me back up one more time. That silly troll thread started out typically as all comparison threads in forums do. Some, like me, said the Dreamcast was probably good enough for the time and had great image quality, the majority even in this Sega forum said that the PS2 was obviously the most powerful system with dramatic increases in polygon performance in particular. Did the games show this? If so were they the exception, the mean or median? When were these games released? .

Yet you didn't just post this part I bolded, you went on to post yet more of your anti Playstation/Sony ranting, probably expecting that people would enthusiastically applaud you since you are on a Sega forum where its cool to hate on the Playstation. However this time you had people pushing back against your claims, the sad thing is that people aren't attacking you personally like you are trying to portray, they are merely refuting your views, which as far as I know is part of how a forum functions.
If you had just said that you felt the Dreamcast was good enough and left it at that I expect that everything would of been fine and dandy. Oh BTW if you felt that the thread was such a "silly troll thread" why even bother posting in it?

Obviously you associate me with ABF and others, which is why you persistently reply to flames against me with "you need to spread around more rep... bla bla" It's like you can't read a single thing I write without seeing it as something horrible against something you find great. I can understand that, the PS2's marketing was quite successful at establishing it as the most powerful system in the eyes of its supporters..

Is this where you try to portray me as attacking you personally again? What I said was that this ONE statement came across to me as very fanboyish and sounded remarkably like the crap ABF was/is posting. As for my choices of repping posts (or mostly attempting to rep) I normally BOLD the part I am mostly agreeing with none of which are what I consider "flaming" most of them came across to me as attempting to ask you to dial down the Sony hate a few notches, in none of those posts has Barone called you names or made personal attacks on you but I guess with you its a case of "You either with me or you're against me - no in between"

In the light of this quite prolific belief, actually no I don't even need to put this under that expectation. When I see a 224/240p PS2 game or even a, what 448 line PS2 game with 4-bit textures in most places and no anti aliasing to speak of I do, in fact, think it looks like ass. I'd rather play a PS1 or Saturn game than a PS2 game with those kinds of graphics. When the PS2 runs at a resolution more in line with reasonable 6th generation expectations I think it looks fine. The problem is, in my not inconsiderable experience with the PS2 library since 2000, including three years of retail experience, those games that look like they should are far, far, far fewer than Sony advocates would ever want to admit..

And in certain cases I would actually agree with you, Ridge Racer 5 is one of the games that I would label as not being much better than a PS1 game graphically. Just an FYI for you I'm not the massive PS2 fanboy as you are trying to portray me as, in fact I was pretty disappointed in the PS2 at first so disappointed in fact that I really wasn't interested in the PS2 and moved towards PC gaming for much of the 6th gen, I got my PS2 in 2006 mainly because it was cheap and because the games were cheap. Having said that though I don't have such an extremely negative opinion of the PS2 as you obviously do and I don't believe it deserves the relentless trashing you (and others) give it.

All of that is coming around to say, the very common claim that the PS2 is "way more powerful" really does not show up in most games but people believe it because that is how marketing works. The same people also tend to claim the Dreamcast was already maxed out in 2000-01, and that its polygon performance was so low that it was dead as soon as the PS2 launched. If that were so, then most PS2 games would have been laughed off the shelf for their "poor" performance.

Well like I said you might well run in to that kind of opinion on a Sony forum, BUT you aren't on a Sony forum, are you?

[sheath]

You know, you aren't posting on some hardcore Playstation site where they they mindlessly believe EVERYTHING Sony says, I already know that most of Sony's Hype was bullshit and I wasn't specifically referring to what Kutaragi claimed. I was more referring to where you said "Why? The entire generation was floating around 1 Million polygons per second, especially outside of Racers and FPS." you were quite clearly indicating that you believed that since the the entire generation never got above 1 million polys/sec that the Dreamcast could quite easily keep up with the PS2, when it has already been proven that this 1 million polys/sec max is untrue, yet instead of accepting this and moving on, you attack the people who are refuting your claims by accusing them of "flaming and trolling" you. How about this quote from you? "The bottom line is the Dreamcast was more capable of surpassing 3 million polygons per second than the PS2 and the entire generation floated around 1 million with a few exceptions on each platform." That is plainly incorrect, are you now stating you were wrong in that claim?

Uh, yes, that later quote is plainly incorrect. I'm not even sure why I would have said that. I know why I was stating so plainly about the entire generation floating around the 1 million mark, that was based on Resident Evil 4 PS2. I was also sick of the myth around the PS2's polygon prowess constantly cropping up in any 6th generation discussion without any evidence, such as the developer quotes that have now been brought to the table.

-edit-
I just remembered why I said the second statement above, it was before the VUs being designed to stream polygon data to the GS instead of polygon data needing to be stored in VRAM was brought up. If the GS's 4MB of VRAM was to include polygon data the Dreamcast PVR2-DC would have been more capable of surpassing 3 million polygons per second. Since the VU1 can stream more polygons than that by quite a bit per second this point is wrong. This is exactly the kind of info I have always hoped this thread would bring up.
-/edit-

We aren't going to go on about everything I have ever written on the internet having to be right in order for me to have the right to post today are we?

Yet you didn't just post this part I bolded, you went on to post yet more of your anti Playstation/Sony ranting, probably expecting that people would enthusiastically applaud you since you are on a Sega forum where its cool to hate on the Playstation. However this time you had people pushing back against your claims, the sad thing is that people aren't attacking you personally like you are trying to portray, they are merely refuting your views, which as far as I know is part of how a forum functions.
If you had just said that you felt the Dreamcast was good enough and left it at that I expect that everything would of been fine and dandy. Oh BTW if you felt that the thread was such a "silly troll thread" why even bother posting in it?

Did you notice I stopped? What I'm still dismayed about is how serious it got because I posted a few videos of various similar games on both platforms. That and how people took those videos as me saying the Dreamcast was positively absolutely unquestionably the best in the comparison. As for what I expect from people for my views, I'm leaning toward more self-promoting cynicism and a poor reading of my actual statements.

Is this where you try to portray me as attacking you personally again? What I said was that this ONE statement came across to me as very fanboyish and sounded remarkably like the crap ABF was/is posting. As for my choices of repping posts (or mostly attempting to rep) I normally BOLD the part I am mostly agreeing with none of which are what I consider "flaming" most of them came across to me as attempting to ask you to dial down the Sony hate a few notches, in none of those posts has Barone called you names or made personal attacks on you but I guess with you its a case of "You either with me or you're against me - no in between"

If you start telling me what my words mean and not allowing me to A) admit I was wrong and B) reiterate what I said to add clarity I will see that as a personal attack. It is pure arrogance to do so, but quite common among forum goers, along with poor reading comprehension and zero personal accountability.

And in certain cases I would actually agree with you, Ridge Racer 5 is one of the games that I would label as not being much better than a PS1 game graphically. Just an FYI for you I'm not the massive PS2 fanboy as you are trying to portray me as, in fact I was pretty disappointed in the PS2 at first so disappointed in fact that I really wasn't interested in the PS2 and moved towards PC gaming for much of the 6th gen, I got my PS2 in 2006 mainly because it was cheap and because the games were cheap. Having said that though I don't have such an extremely negative opinion of the PS2 as you obviously do and I don't believe it deserves the relentless trashing you (and others) give it.

I honestly don't know what you think aside from that you disagree with some run on of anything I've ever said and you like it when somebody posts against me. As for my "relentless trashing" of the PS2, it is worth noting that I only respond with negatives in threads I was already involved in, or technical threads, when the consensus seems to be the PS2 is flawless and uber powerful. It's not like I go looking for these threads either, any topic that gets into the PS2's technical capabilities must end in its supreme awesomeness and anybody opposed is just an ignorant fanboy. That is how over the top slavishly loved the PS2 is even today, even in this group aside from a handful of argument factory trolls and apparently me.

Well like I said you might well run in to that kind of opinion on a Sony forum, BUT you aren't on a Sony forum, are you?

Seems like it is just as prevalent here as anywhere else, the only difference is these threads had a semblance of respectful discussion to them at first (not the graphics thread, that was stupid from the start).

[Barone]

Any idea about what actually was the Ubisoft's Geotexture (they hyped about it back then)?

Not a clue. Might have been mega texturing where they splatted a whole load of textures into VRAM on the fly and concatenated the draw calls into one by for world geom. I'll have to look it up.

I found this:


And this:
"More importantly, a newly created technique, dubbed "geo texturing," has allowed Chaos Theory to mimic the normal mapping on the Xbox using a real-time 3D mesh that's textured on the fly. The end result is the same look as the Xbox's normal maps, done in real 3D, which doesn't tax the PS2 as much as traditional techniques would. The cherry on top of the whole package is a stunning new water effect that is quite possibly the most stunning water we've seen on the PS2 to date--yes, even better than the water in Splashdown and Baldur's Gate. The effect is actually more dynamic than what we've seen on the PC and Xbox versions and features waves and ripples that naturally react to movement."
http://www.gamespot.com/articles/tom.../1100-6116458/

[stu]

Uh, yes, that later quote is plainly incorrect. I'm not even sure why I would have said that. I know why I was stating so plainly about the entire generation floating around the 1 million mark, that was based on Resident Evil 4 PS2. I was also sick of the myth around the PS2's polygon prowess constantly cropping up in any 6th generation discussion without any evidence, such as the developer quotes that have now been brought to the table.

We aren't going to go on about everything I have ever written on the internet having to be right in order for me to have the right to post today are we? .

Um I really have no interest (and certainly not the time) to go trawling through every post you have ever posted here to merely nit pick where you have posted incorrect information, I was merely pointing out that a lot of the argument started with claims that the PS2 really was no better than the Dreamcast and that when people attempted to prove that incorrect it was met with accusations of flaming and trolling, where as a simple "My bad I was wrong" would of resolved it. As for your right to post I have no power to decide that and I think its kinda silly to suggest it.

Did you notice I stopped? What I'm still dismayed about is how serious it got because I posted a few videos of various similar games on both platforms. That and how people took those videos as me saying the Dreamcast was positively absolutely unquestionably the best in the comparison. As for what I expect from people for my views, I'm leaning toward more self-promoting cynicism and a poor reading of my actual statements.

I'm sorry but from your 1st post you were already posting negatively regarding the PS2, from being overly critical (imo) of the controller to making it out to be a glorified DVD/PS1 player, you are of course entitled to your opinion and you explained later that this was how you thought about the PS2 when purchasing it but others are also entitled to disagree with it and say so and for me personally it kinda set the tone.

If you start telling me what my words mean and not allowing me to A) admit I was wrong and B) reiterate what I said to add clarity I will see that as a personal attack. It is pure arrogance to do so, but quite common among forum goers, along with poor reading comprehension and zero personal accountability.

Um I think you are over reacting here a tad tbh, speaking personally I was interpreting your posts and felt that they focussed a lot on the negatives of the PS2 and being highly critical of games that only mainstreamers buy and I told you that. I don't feel that I have "poor reading comprehension" and I don't believe anyone else in that thread who was, in your view, "trolling and flaming" you has any problems interpreting your posts either.

I honestly don't know what you think aside from that you disagree with some run on of anything I've ever said and you like it when somebody posts against me. As for my "relentless trashing" of the PS2, it is worth noting that I only respond with negatives in threads I was already involved in, or technical threads, when the consensus seems to be the PS2 is flawless and uber powerful. It's not like I go looking for these threads either, any topic that gets into the PS2's technical capabilities must end in its supreme awesomeness and anybody opposed is just an ignorant fanboy. That is how over the top slavishly loved the PS2 is even today, even in this group aside from a handful of argument factory trolls and apparently me.

Tbh I went over the 1st couple of pages of that thread and I only counted one or two posts that I would class as being even remotely close to praising the PS2, however your 1st post left the reader in no doubt how you felt about the PS2 and kinda set the tone for the discussion.
As for me disagreeing with everything you have ever said I think I have been fairly rigid in my views and haven't merely flip-flopped my view to oppose you for the hell of it, mostly when I have posting dissenting views from you it is when you are posting negative appraisals of the Playstation console and it's games and also on the subject of the 32X, I feel I've been respectful and have given my reasonings in a clear and concise manner. I respect your right to your opinion and expect to be able to give mine even if it is in disagreement, I've covered my repping of posts in my previous response to you and will only say that I don't rep those posts with the view of "ganging up" on you but because I either agree with all or some aspect of what the poster has said.

Finally on the subject of the general consensus of that thread I felt that the general consensus was the the Dreamcast was up to the task of keeping up with the PS2 for a considerable amount of the Dreamcast's lifespan, had Sega been in the position to continue, however as the PS2 started to get tapped more efficiently and the games utilized more of it's potential that some cut backs may have been necessary for any potential later Dreamcast ports. (probably beyond 2003/4)

Seems like it is just as prevalent here as anywhere else, the only difference is these threads had a semblance of respectful discussion to them at first (not the graphics thread, that was stupid from the start).

I really don't see it personally, if anything there appears to be way too much "Nintendo Worship" considering this is a Sega forum, but meh I guess thats just me.

[rusty]

I found this:

And this:
"More importantly, a newly created technique, dubbed "geo texturing," has allowed Chaos Theory to mimic the normal mapping on the Xbox using a real-time 3D mesh that's textured on the fly. The end result is the same look as the Xbox's normal maps, done in real 3D, which doesn't tax the PS2 as much as traditional techniques would. The cherry on top of the whole package is a stunning new water effect that is quite possibly the most stunning water we've seen on the PS2 to date--yes, even better than the water in Splashdown and Baldur's Gate. The effect is actually more dynamic than what we've seen on the PC and Xbox versions and features waves and ripples that naturally react to movement."
http://www.gamespot.com/articles/tom.../1100-6116458/

The video isn't working for me. But it's either some sort of procedural texture technique or tessellation, or both. It wouldn't surprise me if this is the case, because you could do both on the VS and GS without any real stoppers.

[sheath]

I found a few more technical documents I'm sure everybody here has already seen, memorized and otherwise digested. Posted here because I have been using these threads to slowly collate the available info into a bigger comparative picture (if alternative skepticism prevails please see sig).


http://www.philvaz.com/games/PS2.htm (Archive.org earliest record date March 4, 2011)

Abstract/teasers:
"

Benefits of A Micro-programmable Graphics Architecture
By Dominic Mallinson

Introduction

During this session, I want to examine the benefits of a micro-programmable graphics architecture. I will consider the graphics pipeline and where hardware and software techniques can be applied. After looking at the pros and cons of hardwired hardware vs. CPU vs. micro-coded coprocessors, I will discuss procedural vs. explicit descriptions. Finally, I hope to demonstrate some of the principles by showing examples on a specific micro-programmable graphics system: the Playstation 2."

...

"

PS2 : Vector Units
The vector coprocessors are SIMD floating point units. They perform multiply/accumulate operations on 4 single precision floats simultaneously with single cycle throughput. In parallel with the FMAC operations, the vector units perform single float divide, integer and logic operations.
PS2 : Vector Unit 0 (VU0)
The VU0 has 4K of instruction RAM and 4 K of data RAM.
This unit is closely coupled to the CPU and can be used as a MIPS coprocessor, allowing the CPU instruction stream to directly call vector unit instructions.
The VU0 can also be used as a stand-alone, parallel coprocessor by downloading microcode to the local instruction memory and data to its data memory and issuing execution instructions from the CPU. In this mode, the CPU can run in parallel with VU0 operations.
PS2 : Vector Unit 1 (VU1)
The VU1 has 16 K of instruction RAM and 16 K of data RAM.
This unit is closely coupled to the Graphics Synthesizer and has a dedicated bus for sending primitive packet streams to the GS for rasterization.

The VU1 only operates in stand-alone coprocessor mode and has no impact on CPU processing which takes place totally in parallel. Downloading of microcode, data and the issuing of execution commands to VU1 are all accomplished via the DMA Controller."



...

"
Pros and Cons of Micro-Programmability
When compared to a fixed hardware implementation, micro-programmed hardware has the following pros and cons :
Pros : (Micro-program vs. Hardwired Hardware)
Flexibility and generality. Slight variations and parameterizations of an algorithm are easy. A wider set of needs can be addressed.
Tailored (and optimized) to application and data set whereas hardwired hardware tends to force application functionality to fit the available hardware.
Multiple use. Micro-programmed hardware can be tasked with performing very different tasks by using a different set of microcode.
Can reduce bus-bandwidth. Hardwired solutions typically require a fixed data format to be passed in. Often this is more verbose than a specific object or application requires. Microcode allows the data set to be passed to the hardware in a more "native" format. This typically reduces the amount of data passed across the bus from main memory to the graphics coprocessors.
Non standard effects. e.g. non-photorealistic rendering, distortion, noise, vertex perturbation etc.
Cons: (Micro-program vs. Hardwired Hardware)
Usually slower. Typically, dedicated hardware can get better performance than microcoded hardware. However, this is not always the case. Its worth noting that certain combinations of graphics algorithms share intermediate calculations and data. In a hardwired solution it may be necessary to redundantly repeat these intermediate calculations whereas a microcode implementation may recognize these special cases and optimize appropriately."

http://segatech.com/technical/dcblock/index.html (Archive.org earliest record date March 6 2001)


"

Data Path	Bandwidth
CPU SH-4 to 16 MB Main Memory	64-bits x 100 MHz = 800 MB/s
PVR2DC to 8 MB Graphics Memory	64-bits x 100 MHz = 800 MB/s
ARM7 Sound to Sound Memory	16-bits x 66 MHz = 132 MB/s
Total:	1732 MB/s

"
External Bandwidth
Controller Ports: These ports have a rated bandwidth of 2 Mbits/s, or 250 KBytes/s. This is 1/6th the speed of a PC USB (Universal Serial Bus) port, which can transfer data at 12 Mbits/s. Even though it is quite a bit slower then USB, it still more then fast enough for any data transfer that has to take place between the controllers and the main unit. A 128 KB VMS memory unit can transfer it's entire memory contents to the Dreamcasts main memory in half a second.
The port coming off of the SH-4 is the rear serial port, that has the same performance characteristics as the controller ports, which is about 2 Mbits/s, or 250 KBytes/s. This port can be used for networking two DC's together, or attaching external devices like a keyboard, etc.
Modem Port: This port is attached to a bus that is rated at 16-bits x 25 MHz = 50 MB/s, and this same bus is also used to transfer sound data from the GD-ROM to the sound system. The speed of this port clearly shows that it cannot support any processor or memory expansion at all, which is contrary to what others have suggested. It seems quite clear by this diagram that the Dreamcast has no external ports that can support a processor or memory expanision, as the ports are not fast enough. "

Examples of the above disproven contemporaneous assertion:
http://www.shinforce.com/dreamcast/Expandability.htm (Archive.org earliest record date November 18, 2000)
(Placeholder for Sega press releases)
Naomi GD-ROM, 32MB RAM, 16MB VRAM,
Naomi 2 GD-ROM 2X SH-4, 2X PVR2, Custom Videologic T&L "Elan" 100Mhz

Other discussion on Dreamcast expansion:
http://dreamcast-talk.com/forum/viewtopic.php?t=4532

[Crazyace]

Interesting Shenmue postmortem here:

http://www.gamespot.com/videos/class.../2300-6417813/

Yu Suzuki marks the Dreamcast as 3M polygons/second peak and 1.5M polygons/second expected in game. ( Around 18.59 in the video )

[zyrobs]

Interesting Shenmue postmortem here:

http://www.gamespot.com/videos/class.../2300-6417813/

Yu Suzuki marks the Dreamcast as 3M polygons/second peak and 1.5M polygons/second expected in game. ( Around 18.59 in the video )

More importantly he outright states that those are lower than the final hardware specs because it was just what they predicted what it will likely have.

[sheath]

I would be interested to know what Sega considered a polygon when estimating the Dreamcast at 3.5 million peak in game. Since we have already discussed the many definitions of polygons, including multipass effects and peak versus average rates per game, I think it is now necessary to know exactly what kind of polygons Suzuki was talking about. Textured and lit with one light? Lit with three lights? Just textured with no lighting and burned in shadows? Just textured and gouraud shaded with only modifier volume shadows? Textured with 256x256 size textures or smaller?

Most importantly, since the Dreamcast only renders in 32x32 tiles, does that mean that the Dreamcast's 3.5 mpps "peak" is actually 3.5 million 32x32 tiles of front facing non obscured textured polygons? In other words does the back facing and obscured geometry of ~40 bytes per polygon effectively go away, or is it still in VRAM until final rendering? I would imagine that in order for the tile based renderer to be effective that it simply does not see those back facing or covered polygons in the first place, meaning that they aren't even in VRAM at the point of rendering. I just haven't seen that described in detail.

Either way, I think the group has established that 3.5 million Dreamcast polygons are essentially incomparable to 3.5 million or 6 million or 18 million PS2 polygons. I'd like to know for certain that they are incomparable marketing specs with little or no comparable similarities before tossing out the entire polygon discussion Sony started with the PS2 hype steamroller and Microsoft went ahead and perpetuated.

In related news, I just watched a video by a kid who calls himself "the gaming historian." He actually compared the Dreamcast's 3.5 Million polygons to the PS2's 60 million polygon spec as 1:1 obvious and established fact.

[Yharnamresident]

I would be interested to know what Sega considered a polygon when estimating the Dreamcast at 3.5 million peak in game. Since we have already discussed the many definitions of polygons, including multipass effects and peak versus average rates per game, I think it is now necessary to know exactly what kind of polygons Suzuki was talking about. Textured and lit with one light? Lit with three lights? Just textured with no lighting and burned in shadows? Just textured and gouraud shaded with only modifier volume shadows? Textured with 256x256 size textures or smaller?

Most importantly, since the Dreamcast only renders in 32x32 tiles, does that mean that the Dreamcast's 3.5 mpps "peak" is actually 3.5 million 32x32 tiles of front facing non obscured textured polygons? In other words does the back facing and obscured geometry of ~40 bytes per polygon effectively go away, or is it still in VRAM until final rendering? I would imagine that in order for the tile based renderer to be effective that it simply does not see those back facing or covered polygons in the first place, meaning that they aren't even in VRAM at the point of rendering. I just haven't seen that described in detail.

Either way, I think the group has established that 3.5 million Dreamcast polygons are essentially incomparable to 3.5 million or 6 million or 18 million PS2 polygons. I'd like to know for certain that they are incomparable marketing specs with little or no comparable similarities before tossing out the entire polygon discussion Sony started with the PS2 hype steamroller and Microsoft went ahead and perpetuated.

In related news, I just watched a video by a kid who calls himself "the gaming historian." He actually compared the Dreamcast's 3.5 Million polygons to the PS2's 60 million polygon spec as 1:1 obvious and established fact.

But everyone mocks the power of the Dreamcast, hes just trying to fit in.

[Crazyace]

Hi Sheath,

I found some of the original Sega spec documents here ( still cant find the PDF though )

http://www.goodcowfilms.com/farm/gam...EE%20Micro.htm
http://www.goodcowfilms.com/farm/gam...20Page%202.htm

The most relevant parts are as follows:

"The 64-bit-wide, 100-MHz system bus between the CPU and the ASIC can transfer polygon data at up to 8 million stripped triangles per second."
"The final form has an internal floating-point unit of 1.4 Gflops, which can calculate the geometry and lighting of more than 10 million polygons per second."
"Fill rates are a maximum of 3.2 Gpixels per second for scenes comprising purely opaque polygons, falling to 100 million pixels per second when transparent polygons are used at the maximum hardware sort depth of 60."
"Overall rendering engine throughput is 7 million polygons per second, but in Dreamcast, geometry data storage becomes the limiting factor before pixel engine throughput."
"Recent tests by Sega, using the company's graphics libraries proved that Dreamcast uses 100% of the CPU and rendering engine's ability and delivers 6 million textured and lit polygons per second. Although game logic and physics reduce peak graphic performance, the performance of recent 3D titles for the system from many companies clearly demonstrates the platform's graphics superiority."

"Drawing rate greater than a million polygons per second
Fill rate greater than 3.2 Gpixels per second"

I would be interested to know what Sega considered a polygon when estimating the Dreamcast at 3.5 million peak in game. Since we have already discussed the many definitions of polygons, including multipass effects and peak versus average rates per game, I think it is now necessary to know exactly what kind of polygons Suzuki was talking about. Textured and lit with one light? Lit with three lights? Just textured with no lighting and burned in shadows? Just textured and gouraud shaded with only modifier volume shadows? Textured with 256x256 size textures or smaller?

I expect in most of the tests above a polygon is a simple gouraurd textured triangle - small sized so the pixel fill rate isn't the bottleneck. The peak transform rates are without any lighting.

Most importantly, since the Dreamcast only renders in 32x32 tiles, does that mean that the Dreamcast's 3.5 mpps "peak" is actually 3.5 million 32x32 tiles of front facing non obscured textured polygons? In other words does the back facing and obscured geometry of ~40 bytes per polygon effectively go away, or is it still in VRAM until final rendering? I would imagine that in order for the tile based renderer to be effective that it simply does not see those back facing or covered polygons in the first place, meaning that they aren't even in VRAM at the point of rendering. I just haven't seen that described in detail.

DC needs all of the polygons stored in vram, along with lists of which triangles intersect each 32x32 tile. ( A of 20x15 lists for a 640x480 screen ) it then goes through and draws those polygons. In some cases twice the storage in vram is needed - 1 buffer for the drawing, and another buffer for the tiling of the next frames polygons.

Either way, I think the group has established that 3.5 million Dreamcast polygons are essentially incomparable to 3.5 million or 6 million or 18 million PS2 polygons. I'd like to know for certain that they are incomparable marketing specs with little or no comparable similarities before tossing out the entire polygon discussion Sony started with the PS2 hype steamroller and Microsoft went ahead and perpetuated.

There are 3 sets of specifications around... the first comes from the engineering / press specs. ( Above article for Dreamcast )

DC: 10M transformed on CPU, 8M transferred to gpu memory,7M rendered
PS2: 37.5M transformed ( ~7 cycle VU1 vertex program ) , 37.5M drawn

DC: 3.2GPixels Z test , 100MPixels drawn. ( Z test occurs before texturing on DC )
PS2: 3.2GPixels non textured, 1.6GPixels textured.

The 2nd sets come from the various game engine figures in the press

DC: 5M ( Melbourne house Test drive )
PS2: 18M ( Melbourne house Grand Prix Challenge )

The 3rd set comes from performance captures ( DC emulator, or PS2 PA kit via presentations )

DC: ~600K vertices/second for Sha Hua in Shenmue Passport disc captured on Null DC 1.04
DC: ~2.4M vertices/second maximum seen during DOA2 special edition attract sequence , again captured on Null DC 1.04
PS2: 6.21M polygons/second for some Ratchet Game ( slide 46 and 47 of http://lukasz.dk/mirror/research-sce...er_18Mar03.pdf )

[sheath]

Thanks for the breakdown, I only looked for people trying to break down the counts with emulators briefly and didn't find anything. Earlier in this thread it was determined that the maximum the PVR2-DC could set up was 4.2 million polygons per second. I forget what the logic was exactly but it had something to do with the frame buffer plus 40bytes per polygon. Test Drive Le Mans' 5 million figure was also dismissed as incorrect based on a couple of NullDC vector gathering attempts on another forum. I didn't book mark exactly where in the thread it was discussed but will try to find it later today.

In the Ratchet sample shown in pages 46-47 of the Performance Analyzer Doc, it says primitives kicked on the row with the 6.21M polygons. Is that jargon for displayed polygons?

[evilevoix]

So from what I'm reading the PS2 wipes the floor with the Dreamcast. It's not even close. What a shame.

[sheath]

So from what I'm reading the PS2 wipes the floor with the Dreamcast. It's not even close. What a shame.

Well, with a BIOS update in the GD-ROM the Dreamcast was in the same generation.

The Ratchet game above is showing 6 million polygons per second and is considered technically advanced for the PS2. Melbourne House's interview about Grand Prix Challenge says that "most" other games were doing 6 million polygons per second max compared to their 12-13M, or 18M with "multipass effects". But yes, I think most Dreamcast games get trounced technically by some PS2 games, especially after the performance analyzer was released to third parties in 2003.

The Dreamcast never would have had such a tool you see, it isn't possible to improve Dreamcast games in any way. It was completely maxed out by like two games.

Seriously though, I think both systems have a fairly steep list of caveats for maximum performance to be achieved. For the Dreamcast the developer had to focus enough on the platform to optimize for the tile based renderer and not use Windows CE and not port from the PS1 or N64. The PS2 needed the developer to be savvy to Assembler level parallelism and the game could_not consider multiplatform needs. From what I can tell based on developer interviews both conditions are likely to produce more "bad ports" and poorly optimized games than not. And here we are.

[Crazyace]

The 6M polygon figure for that Ratchet capture is 'drawn by the GS' , so compares pretty directly with the figures shown by NullDC for Dreamcast games.
As a comparison NullDC shows around 300 thousand polygons/second at most times for Test Drive , with the highest count around 1.2 million during the intro for the 24 hours race. As the picture for the Ratchet capture looks like a quiet area ( no enemies close to the main character ) you cant assume that it represents the highest poly count - it may be more comparable to the 300k figure.
Also for Test Drive ( and in general ) both DC and PS2 used "multipass effects" , so the 6M engine peak should compare directly with 18M.