MiniFAQ : How CELL works.

[zurich]

Yeah. It's a pity there isn't a PA for other systems to expose the wild exagerations (or just over optimistic profiling) that goes on for them too...

I suspect that most of the people making bold claims would be equally surprised by PA data.

You an get stat's directly from the GPU on Xbox, including Pixels drawn and triangles processed. You don't get any nice bus usage stats but those are less useful in this context. GC has similar built in performance counters.

And FWIW we weren't at all surprised by our PA results. There is some question as to what is classified as a triangle by the PA (i.e. at what point in the pipeline it's measuring), but from a dev standpoint the total tri count isn't really useful data.

When you're doing the analyzing part, does the game run a rolling benchmark type of thing? Does someone stand there and play through a level or something? Always been curious about that.

 

[chaphack]

So what are the average PA-ed polygon count of a PS2 game? 6-8mpps?

 

[cthellis42]

Probably even lower if you count ALL PS2 games. Heh... After all, there's a ton that need no real effects, no real power, no real speed...

But if you're excluding the games that would have a tough time taxing a Pentium II with a Voodoo2, you're probably in that ballpack somewhere.

 

[ChryZ]

When you're doing the analyzing part, does the game run a rolling benchmark type of thing? Does someone stand there and play through a level or something? Always been curious about that.

Both would work, but a "rolling benchmark" for the sake of glory is not very common or do you remember any developer/publisher bragging with their PA results. The PA is more a developer tool, take it as a sign, that there IS a Sony developer support

 

[zurich]

When you're doing the analyzing part, does the game run a rolling benchmark type of thing? Does someone stand there and play through a level or something? Always been curious about that.

Both would work, but a "rolling benchmark" for the sake of glory is not very common or do you remember any developer/publisher bragging with their PA results. The PA is more a developer tool, take it as a sign, that there <B>IS</B> a Sony developer support

Ya, but I'm asking what they test? They have to run SOMETHING through the PA.. is it a benchmark, rolling demo..?!

 

[Fafalada]

and I'm pretty sure after a PA session with those games you will not be so sure anymore

Of course it also doesn't help that PA context of numbers is quite different from what people normally measure in their engines on any platform (ie. primitives submitted to GPU vs primitives actually rasterized).

And FWIW we weren't at all surprised by our PA results. There is some question as to what is classified as a triangle by the PA (i.e. at what point in the pipeline it's measuring), but from a dev standpoint the total tri count isn't really useful data.

It's counting only primitives that reached DDA stage. It also doesn't help that it counts all primitives equally - throwing around a couple of thousand non-triangle primitive to get optimal fillrate use during framebuffer operations will definately register on that total number.

 

[ChryZ]

They have to run SOMETHING through the PA.. is it a benchmark, rolling demo..?!

Any executable code should work, e.g. rolling demo or someone playing a build of game in development. The PA listens to the hardware and collects analyzable data, while the code is executed.

 

[ERP]

The PA just grabs several ms' of data when a foot switch is pressed, could be rolling demo, could be be gameplay. IMO the later is of more use.

It's counting only primitives that reached DDA stage

The PA manual is completly unclear on whether it counts degenerate/scissored or skipped tris, doesn't really matter, I just would have expected a clearer description in the manual. The numbers only really useful from an optimization standpoint when your looking at portions of the scene anyway.

 

[London Geezer]

I would have thought they would want to test their games under gameplay environment. after all they need the PA to check how the game is performing and therefore how playable it is (frame rate issues etc). what would be the point of running a rolling demo if not for research?

I MEAN, i dont think the first priority of a programmer would be "to see how many polygons my game is pushing", rather "let's see how i can get rid of that bit of slowdown HERE and.... THERE..."
see my point? still, i'm no programmer so what do i know...