PS2 performance analyzer statistics (from Sony)

[McFly]

http://www.technology.scee.net/sceesite/files/presentations/PSP/HowFarHaveWeGot.pdf

What a shame that there are still many developers that are not even using VU0.

Fredi

 

[function]

Interesting! [The pdf as a whole I mean, not just this polygon stuff]

So the "fastest so far" is coming in at 7.5 million polys a second, with most running between 2 -5 mpps.

It's quite a bit different from the 15 - 20 million figure that has been used by the console's advocates (infact this would seem to roughly indicate that the DC was to PS2 what the PS2 is to Xbox in terms of performance).

As has been mentioned sereral times on these boards, the PS2's CPU does appear to be the weak point of the system.

 

[passerby]

Nice find.

What a shame that there are still many developers that are not even using VU0.

Any confessions from the resident devs?

But I too, am quite surprised at the slide which says "Best performing games use up to 8% VU0". Not sure if they scanned the just released J&D2, R&C2, or SH3 and ZoE: Anubis from Japan. Because that statistic is really surprising.

 

[nAo]

We use VU0 to perform skinning and to accelerate physics...its use depend on the stuff that we manage at any given frame but it's rare VU0 utilization goes up beyond 20% of a PAL frame time.

ciao,
Marco

 

[McFly]

As an 80% artist 20% coder myself (not for PS2):

- Use palletised textures when possible
- More friendly to the DMA and VRAM
- Good quantiser is essential
- Swizzle for optimal performance
- Don't listen to the artists, convert the textures yourself
- Don't tell them I said that

I have to say that I never listen to the coder (even if it's me). It's better anyway if you do high res textures and the coder decides what res or color depth he can work with or you would have to do them again if the coder finds a way to use higher res textures.

So in my case, tell 'em!

Fredi

 

[Squeak]

It’s a bit depressing that they (SCEE) still have to emphasise the same issues, they did three years ago.
Are the developers stupid? Or is it really that incredibly hard? Or do they have to keep the datasets and code general, to make multiplatform games possible? No matter what, it’s a shame that PS2 is still under 50% utilized.

 

[Deadmeat]

...

This is a troubling statistics indeed.

- 2% VU0 usage : Most games are still not using VU0.
- 2.3 MB of data sent to VIF1 for geometry.
- 1.5 MB of data sent to GIF for texture.

I already predicted a poor utilization rate like this within days after Kutaragi announced PSX2 back in 1999, it was so obvious. In the end, the PSX2 was no better than DC in sustained performance(This is why DC and PSX2 games don't look all that different) but the marketing hype did the DC in.

Now consider the implication of this performance analysis. If SCEI coders are having so much trouble properly untilizing just two VUs after 5 years of exprience, how the hell do you expect them to deal with 16 APUs in PSX3???

A hardware must be designed to serve software's need, not the other way around. Now do you see why I criticize poor PSX2 & PSX3 designs so much???

 

[Deadmeat]

...

Are the developers stupid?

Nope.

Or is it really that incredibly hard?

Blame the machine. Okamoto admitted that PSX2 was one of the craziest design in computer science history. You can't make a puppet dance when the strings are uneven.

It's quite a bit different from the 15 - 20 million figure that has been used by the console's advocates (infact this would seem to roughly indicate that the DC was to PS2 what the PS2 is to Xbox in terms of performance).

These people are clueless. When even the most well-informed forks like Panajev doesn't understand the basics of programming, the rest of so-called "advocates" need not be mentioned.

 

[llemarie]

As the author of the document, I would like to make a few comments:

What a shame that there are still many developers that are not even using VU0.

Very true, but VU0 is not as easy and staightforward to use as VU1. There are many reasons why this is so. We can't condemn developers for that, I am just trying to encourage them to give it a second thought.

It's quite a bit different from the 15 - 20 million figure that has been used by the console's advocates (infact this would seem to roughly indicate that the DC was to PS2 what the PS2 is to Xbox in terms of performance).

Let's not get confused between "top speed" and "average speed". A Porsche can go over 200km per hour, but drive it over hundreds of kilometers and look at the average speed you got.
There are regulations and constraints, there are pauses and traffic jam.
It's all the same on PS2. You can achieve up to 22MP/s, but the average will still be around 5MP/s.

As has been mentioned sereral times on these boards, the PS2's CPU does appear to be the weak point of the system.

True, although not the CPU itself, but rather the CPU efficiency. It is slowed down by bus accesses and stalls, etc... That makes it the weak point.

But I too, am quite surprised at the slide which says "Best performing games use up to 8% VU0". Not sure if they scanned the just released J&D2, R&C2, or SH3 and ZoE: Anubis from Japan. Because that statistic is really surprising.

I did indeed, and the stats do includes the results from those games.


It is not an easy platform to program for, but definitely a very interesting architecture and it does allow for interesting techniques. It is nice to see that it hasn't reached the end of its life yet.
If you have any more questions about the document, please let me know.

Lionel

 

[jvd]

So it looks like its putting out a little more than double what the dreamcast was putting out with what 3 almost 4 more years of dev time ?? Not as impressive as the 20 million polygon number flying around before.

Makes my crazy idea of metal gear being done on the dc not so crazy huh .

 

[Panajev2001a]

I was under the impression ( after talking to some PlayStation 2 developers ) that the Performance Analizer polygon counts are not the ones that are Transformed and Lit by the VUs, but the ones sent to the GS and things like backface culling would reduce the PA score.

Also the VU0 when used as co-processor ( co-op pipe, macro mode ) is not registered as VU0 utilization rate in the PA numbers, but micro mode usage of VU0 ( independent processor, the way VU1 is used ).

 

[Sonic]

Hi, welcome to the board. Thank you for clarifying some of the questions we've had regarding this report. It was very informative.

 

[wazoo]

I applaud Sony for being honest about the real performance of the ps2 games. Any other source would have given this info, it would have been dismissed as pure liar.

I think a corresponding PA analyser for DC would have shown much lower performance for DC games. Remenber that the same devs claiming that they put more than 3M pol /sec on the Dc did also claim that they pushed 18 M pol/sec on the ps2, something this presentation is clearly saying it is wrong.

 

[_phil_]

*No more personal insults . I did not hear deadmeat say anything bad to you for you to say that about him*

 

[Fafalada]

This is a troubling statistics indeed.

Some of it yes, however what you quoted...

"- 2% VU0 usage : Most games are still not using VU0."
VU0 is very much used - in macro mode which doesn't register on PA. Yes, that's not as efficient as micro mode, but it's at least as good as straightforward SH4 FPU use you're pretty fond of.
So it's used to the point you concede to be reasonable (granted not to the point it could be).
"- 2.3 MB of data sent to VIF1 for geometry."
This is enough for 170k-270k vertices per frame, depending on compression used. I am not sure I would call 200+k average troubling. And that's without counting VU actually generating stuff as well.
"- 1.5 MB of data sent to GIF for texture."
Ok here I need YOU to explain Me, why that's supposed to be so bad?

Lionel,
There is one thing I was wondering about - the app with 40x overdraw, is it a big secret which is it?

 

[cthellis42]

Actually, wazoo, the author of said document came on specifically to tell you to NOT make assumptions like that from this information.

Lional, do you mind if I ask where you pull your information from? Individual hand-testing of every game through the PA yourselves only, or is there a lot of developer info-swapping going on as well? Do you test only the maximum performance a title can bring into play, or average it with all the various levels a game can deliver? (Thinking specifically of racing titles here, as you can play with just yourself on a track, or potentially load it up with many more cars--meaning much more needed pushing power.)

I must say your "maximum" count is going to get a lot of people jumping around, but it's supposed to read more like "maximum average polys at 60fps for the entire title," right? Are there actual details you can mention as to what ACTUAL maximums can be hit--or at least held with reasonable stability? Was your 22MP/s giving us a tidbit? There's an extreme dearth of specific information, and broad reports have the tendency to "water down" everything involved, as it's impossible for us to know the full extent without knowing what went into it. (And offhand, I rather doubt the majority of ANY games for any of the consoles push over 5M at this point, as the "good enough" philosophy makes much more money for publishers in general.)

BTW: I loved your comments on the Data Packing slide talking about palletized textures,

- Don't listen to the artists, convert the textures yourself
- Don't tell them I said that

 

[DeanoC]

I don't think its troubling statistics at all.

We are in the last couple of years of PS2's mainstream life. Quality titles have been released this year (RnC2, SH3, etc) showing dramatic improvements from the previous set of games.

We (as devs) have mostly got our head round driving VU1 and GS but finding uses for VU0 is harder. There is still some head room to get some improvement in the next couple of years. The PA will help us find and exploit those areas were were currently weak.

Everything is exactly how it should be.

 

[wazoo]

Actually, wazoo, the author of said document came on specifically to tell you to NOT make assumptions like that from this information.

I agree that the PA is only a way to do performance evaluation, that is why I said that other platforms should be considered with the same angle, then DC games polycount under the same evaluation procedure will be much lower than the 2/3M pol/sec number we have seen in the past for that platofrm.

 

[ChryZ]

Lionel,

the PA actually underestimates the polys drawn, because it doesn't take into consideration, what was rejected by the GS or VU micro-code for drawing, so are stuff like offscreen things or stuff less of the size of a pixel (narrow or small polys).

How does this affect your averages?

You are working for the developer support at SCEE (like listed in you presentation) this results in the following questions:

Will there be something like the PA for the PS3?
Will there be more focus on developer support for future platforms?

 

[function]

It's quite a bit different from the 15 - 20 million figure that has been used by the console's advocates (infact this would seem to roughly indicate that the DC was to PS2 what the PS2 is to Xbox in terms of performance).

Let's not get confused between "top speed" and "average speed". A Porsche can go over 200km per hour, but drive it over hundreds of kilometers and look at the average speed you got.
There are regulations and constraints, there are pauses and traffic jam.
It's all the same on PS2. You can achieve up to 22MP/s, but the average will still be around 5MP/s.

Yeah I'm with you, that's how I interpreted the document. The 15 - 20 million pps performance was being banded around here as an average figure representative of the performance of modern games. The figure of 5mpps fits in perfectly with what Polyphony Digital have been reported as saying about the performance of their Gran Turismo games.

I think it's a perfectly respectable figure too, especially looking at where the DC appeared to be before it died, where the Xbox appears to be now, and how comparable PC components appear to perform. It was with regards to some of the figures being thrown around on this (and other) discussion boards that I was commenting, where I think some people have unrealistically (and unfairly) high expectations of how powerful the PS2 actually is.

But that happens with every console, I'm sure.

I'd like to see similar, detailed figures like these for all of the consoles this generation.