Flay

EA presented this at this year's SigGraph conference: http://www.cs.brown.edu/~tor/sig2002/ea-shader.pdf

The benchmarks are on the last page. The results are interesting. The PC platform is a 1.4GHz Athlon and ATI Radeon 8500.

The first test was a 3998 polygon NBA player:

Gouraud shading, unlit, no textures

PS2: 17.0 MPolygons/second, 22.6 MVertexIndices/Second
Xbox: 47.2 MPolygons/second, 91.4 MVertexIndices/Second
GCN: 18.7 MPolygons/second, NA
PC: 24.1 MPolygons/second, 46.1 MVertexIndices/Second

Gouraud shading, lit, no textures

PS2: 10.9 MPolygons/second, 14.7 MVertexIndices/Second
Xbox: 22.4 MPolygons/second, 43.4 MVertexIndices/Second
GCN: 10.3 MPolygons/second, NA
PC: 15.9 MPolygons/second, 20.9 MVertexIndices/Second

Gouraud shading, lit, textured

PS2: 8.5 MPolygons/second, 11.5 MVertexIndices/Second
Xbox: 14.2 MPolygons/second, 30.3 MVertexIndices/Second
GCN: 7.2 MPolygons/second, NA
PC: 5.1 MPolygons/second, 10.9 MVertexIndices/Second

Then they did a test with a 69451 polygon bunny (from Stanford Computer Graphics Laboratory) with Gouraud shading only:

PS2: 25.2 MPolygons/second, 31.8 MVertexIndices/Second
Xbox: 63.9 MPolygons/second, 93.8 MVertexIndices/Second
GCN: NA, NA
PC: 26.3 MPolygons/second, 36.2 MVertexIndices/Second

Liver&Onions

Let me just say that this doesn't mean jack to me.

And what's with all the NA's in the GC area?

Did they test anti-aliasing?

Flay

Well, 3Dmark folks are certainly used to this. Just some tech info.

An NA beside the GC means it couldn't handle it.

No they didn't test anti-aliasing. Read the PDF I posted if you want to read more.

darkside

Thats all well and good, but why in the heck is there slowdown in Crazy Taxi 3 on the all powerful Xbox?

The Xbox's power advantage is meaningless if the programers aren't doing their job. I like Crazy Taxi 3, but that slowdown shouldn't be there. My concern for the Xbox is not enough games are taking advantage of it's power. Developers need to put the time in to take advantage of what the system can do. Halo is the only game that really shows the system off.

Bruce Lee looks worse that any PS2 game I've seen. It's frustrating to not see the full power of the system used by developers.

It also gives a lot of credit to the arguement that the Xbox power advantage is meaningless. Those polygon counts are not going to make Dead to Rights better than GTA3 on the PS2.

tha_dvd_man

Lazy programming.

Yes, it is all up to the developer. But this is evidence that there is more polygon-pushing potential in the Xbox than the other two systems.

You obviously haven't seen many PS2 games.

agrall

Why didn't they use a more powerful PC? And maybe a GeForce4?

belboz

I just did a quick skim of the paper. It describes a system that EA has apparently been working on for quite some time; probably since before the PS2 was released. Forseeing the era when support for 3 or more platforms was essential to their business model, they've developed a set of tools that allow them to produce multi-platform games more quickly and efficiently. The very short and idealized description of this system is that it takes a single program and data input and compiles it into multiple platform specific, optimized machine code.

They're claiming that the system produces code that runs as fast or faster than any of the hand coded, console specific tools they were using before. I'm not sure if that's saying much, since EA games never struck me as being very well optimized or pushing the envelope when it came to taking advantage of any particular console's hardware.

Anyway, the numbers do give a rough indication of the relative performance capabilities of the various platforms when using this tool. I'm guessing that that NA's for the GC mean that those numbers weren't available or weren't appropriately comparable for whatever reason.

Trigger

I am surprised that the PC fared so poorly against the Xbox. Graphically, they should be more or less equal with that Radeon. Processor-wise, the PC should have a 30% advantage or so. Granted, the PC they used is a budget-type system by today's standards (sub $1000), but was probably closer to top of the line back when the GC and Xbox were released. Also, I skimmed the article and didn't see where they said what OS was being used on the PC... I assume it was Windows 2000 or XP or something since they were talking about DirectX8. That would account for some of those poor showings. I'd also like to know the ram and motherboard chipset. Anyway - proof is in the pudding and my computer generates over 100fps on Quake3 using maxed out graphics options. Xbox runs games at lower resolutions than a PC does and yet a game like Bruce Lee is plagued by slowdowns. Developers are to blame for sure. I'm glad I purchased an Xbox. What's more surprising is that the PS2 outperformed Gamecube. I guess it's more powerful in these circumstances, but GC is able to make the games look better with anti-aliasing techniques and such that the PS2 is lacking in. I'd like to see how my computer stacks up against the Xbox. I'm really quite impressed, but at the same time I know these numbers don't tell the whole story. Anyway - thanks for the article and the info - very interesting indeed.

gcribbs

first problem it that they used a radeon.

the second one is that they used a radeon.

I bet my PC system kills the XBox everyway till sunday.

Had they used a geforce 3 Ti200 card with the better drivers of the nvidia cards I would bet the computer wins.

there is no way a P3 800 wins this over a 1.4 Ghz athlon.

i wonder what chipset they used(didn't see it) or how much memory used.

I would think the console comparisons might be more accurate assuming the code was really opimized for each.

Trigger

I'll bet the new Radeon beats the snot out of your PC and my PC.

belboz

I think the main reason that the XBox fares so well in this test compared to the PC is because its NV2A GPU has two vertex shaders whereas the R200 GPU in the Radeon 8500 only has one.

Also, I think it's important to note that the test was probably run at TV resolutions of 720x480 or 640x480. If you bump up the resolution to those more typical of what PC gamers use, the graphics memory bandwidth quickly becomes the bottleneck and the XBox would be on its knees.

joltaddict

Booooo!!! WEAK!!!

joltaddict

In DEPTH reviews of the specs and marks...

Hardware Behind the Consoles - Part I: Microsoft's Xbox


Hardware Behind the Consoles - Part II: Nintendo's GameCube

ScandalUMD

I looked through the paper, and it seems like EA is just being lazy in their attempts to quickly supply all three platforms.

They are programming graphics to utilize the PS2 graphics synthesizer, and then porting them over. So what this really tests is Flipper vs GS vs the Xchip.

The problem is, the PS2 is designed so that the vector units of the CPU can be used for graphics. The Xbox has a more advanced graphics chip and a less sophisticated CPU. EA seems to be programming games for the PS2 graphics synthesizer, and then porting them to Xbox, failing to properly utilize the capabilities of either system.

Gamecube is left out for some reason, but it's probably not because it couldn't perform the benchmark. It's at least as powerful as PS2, and there's no technical reason for it to be unable to run it. I suspect they simply hadn't ported all the software over to Gamecube yet.

As for the 1.4 Ghz Athlon with the Radeon 8500 getting outperformed, well, that's gotta be EA's fault. Basically, this paper is merely a demonstration of EA's plan to support all platforms with minimal time and money invested in ports. It's not a viable system to system benchmark across all variables.

gcribbs

yep

only the bad drivers are going to hold them back now

joltaddict

The reviews are saying thats not an issue this time around. But it is ATI and I guess history counts for something. Thats one reason Im gonna be waiting, not the only reason and not even that big a reason but a reason nonetheless. A bigger issue is the release dates of the Nv30 and DoomIII. Ill see where they fall.

tha_dvd_man

Maybe they didn't use an absolute top-of-the-line high-end PC because most PC gamers don't own one?

joltaddict

Its not even a Ge3, its an 8500! With an AMD 1.4. Not too many hardcore gamers with that slackass set up.

DigIt

Doesn't matter. As long as it compares to or outperforms the PS2, they're happy.

Flay

The ATI 8500 outperforms the GF3 Ti500 in 3dmark. In fact, only the GF4 series outperforms the 8500.

gcribbs

not in everything. i expect that the specific benchmarking is more slanted towards nvidia chipsets which is why the XBox which is really a lower end pc system is outoerforming the pc they are using.

The athlon 1.4 Ghz is way faster than a P3 700 or 800 that is in the XBox so why would the XBox outperform it??

must be either horrible coding or the radeon or they are using some horrible chipset on the motherboard??

there has to be a reason. they do not give enough information( all the specs of the pc)

frankly maybe they are running the hard drives in PIO mode or something.


if the coding was even decent- that sytem would kill the XBox.

The only other possiblility is that they code for a far slower PC and as a result they are limiting the PC performance themselves.

In which case they should just exclude the PC since this Test tells you nothing.

I do think that the GC should be able to run these tests. the fact that it does not shows that this coding system really sucks

Tamrok

I think there is likely two reasons. One, all the transform and lighting work is done by the NV2A, not the processor, so the difference in processor speeds do not have much, if any, affect on the polygon performance numbers. The cpu is used mostly for physics and A.I. calculations. Two, as someone already mentioned above, the NV2A has an extra pixel and vertex shader (for a total of two each) which is a tremendous advantage because most instructions sent to the vertex shader will require at least two instructions to execute thus speeding up throughput by allowing many shader operations to be completed in a single clock cycle. If those aren't enough reasons, you also have the larger overhead of the Windows operating system compared to the tiny Windows kernel used on the Xbox. Overall, I don't find it the least bit surprising that the Xbox outperforms the PC used in this test. Certainly, the newest graphics cards would have a performance advantage over the Xbox but most PC gamers don't own those cards at the moment. Hence, the reason EA didn't benchmark with a top of the line PC.

gcribbs

I still would :

one like all the specs on the PC(OS,MB chipset, amount of ram...)

two like to see the benchmark with a geforce 3 card which far more people have than a radeon.

my gut is that it is slanted towards nvidia which is why the difference.

Of course hardcore PC gamers do have the top end cards.

My Pc has a Geforce4 4600 to keep it happy with my games

Tamrok

Well, using the Geforce 3, as opposed to the Radeon, may or may not change the numbers slightly but would still ultimately not beat the NV2A (Xbox) because the NV2A is still more advanced than the Geforce 3. The NV2A is basically a Geforce 3 Ti chip with an extra vertex and pixel shader.

Tamrok

That's true. However, EA is not concerned about selling games to only the hardcore PC gamers. They want to sell their games to the widest possible audience. Unfortunately, this results in their games being programmed for a target PC far below the top of the line systems that many hardcore PC gamers own. Thus, the reason why EA used a decent (though not top of the line) system for their benchmark. There are still many PC game buyers who do not even have a graphics card that can match the Radeon 8500, let alone a 1.4 Ghz processor.