I think you're approaching this from the wrong angle. It isn't so much that SPUs are *required*, but that they're better at it than anything else in the system. It's not that you couldn't do without any SPU culling, it's that you have to somehow lighten the vertex throughput requirements in other ways, since RSX is otherwise limited. And yeah, you could do it on the PPU if you really wanted, but even a single SPU will do it significantly faster and because simple things like skinning are easy to turn into several independent jobs, it's more of a "why wouldn't you?" question.
I am not really asking which unit or job distribution is more suitable for backface culling (though it is actually an interesting question by itself). The real question is whether SPU populated CELL has provided any significant performance improvements in terms of gfx yet compared to non-CELL alternatives (before someone objects, I know time time time, hence I say yet).
I have to admit though, I find your comments regarding SPU usage really confusing. Here is what I remember, please correct me if I am wrong:
You said many many times, culling/skinning is easy, even PPE or a single SPUsaid backface can do it.
You said somewhere, EDGE demo was using so many SPUs (each of which can process 750k triangles at 60Hz) because it was a show for "tech people". Now you say "why wouldn't you given that it is easy to parallelize?".
Now it is really obvious that computational complexity of backface culling is linear. I am not proficient in CELL programming but if one SPU is enough for the job I say it is not wise to divide it. As 256k is not enough for geometry, you need to hide memory latency by streaming and it needs to be synchronized properly etc. The last thing I would want to do here is doing other jobs on spare cycles of each SPU as opposed to completely dedicating one to culling. Alternatively of course, CELL's memory bandwidth and latency can be so great(?) I wouldn't care and any trivial implementation would suffice. CELL masters here can clarify I suppose.
You are truly smoking some bad stuff…
R:FOM as far as I can tell doesn’t have an inch of popins or draw distance problems (with nice Multi-sampling to boot) on my TV(s).
Is that so? I suggest you create a game in Manchester (32), walk around south of the map (big opening). Look for barrels, grass, even breakable windows etc. It is really hard to miss unless you are in the middle of heated combat.
Ok just to give you some reference point of view here - a 300MHZ VU unit can do BF culling at a rate of ~500k*60 triangles/sec.
Thanks a lot. That's what I was after. Have you been using culling often on PS2?
And any insight regarding 750k*60 triangles/sec/SPU value which is at least at a significantly higher clock speed?
Don't confuse disagreement with defensiveness. You've posted a thought, and people are disagreeing and saying why. That's the nature of discussion! It's not a moatter of defending a console, but discussing the ideas and trying to open people up to ideas that they may not have considered - both ways.
We are indeed discussing something but I cannot honestly say every one of the replies here is objective (at least on the first page). I even got a negative rep for my initial comment.
If there not using the SPU's for geomrety thats great news, R&C has loads of onscreen activity and geometry so programmed correctly it meens that RSX is more then capable on its own.
That's my thought exactly. Personally wouldn't like to find out first generation games had already been using CELL heavily for gfx.
Give me a break.
