Gholbine said:
The developers and their games have already spoken on this issue: it's a problem, and the generation has yet to begin. Nothing more need be said about it. Take up your arguments with the developers, not me.
OK, fine. You win. This problem is just as bad as N64 vs. PS1
You have a real problem with understanding the context of my messages. There is no definative line between "can" and "need". It's all a grey area. My point is that this is much more grey than N64 vs. PS1. The two scenarios aren't even comparable. You will be very hard pressed to actually see the difference like you could with N64. It's a matter of how many frills the developer wants to and can economically afford to put in, and how much time they want to spend on compacting everything. If you've ever seen the 64KB demos, you'll know there's a plethora of methods for shrinking stuff down.
As for your GPU comments: I stick by what I said. The CPU is not so separated from the graphics as you say. It certainly wasn't in the PS2, and logic dictates that it won't in the PS3 either, and using fabricated knowledge of the RSX to back up your point doesn't exactly work either. Will it help with actual vertex or pixel shading? I doubt it, I never claimed anything like that. You did mention animation, however, which is a massive factor in graphics, and the Cell will certainly help in that department.
Like I said, physics. Animation itself does not need much CPU power at all. It's figuring out what to animate, i.e. physics. And again, the CPU load from physics can be toned down by changing convergence and iteration parameters. I have programmed a physics engine in a 3D game engine, so I have full knowledge in this matter. Even a factor of two CPU power doesn't buy you much, and I doubt that general physics will be nearly that much faster on Cell. From your silence I assume you have zero game programming experience, so your meaningless anecdotes are worthless.
Also, have you seen this demo:
http://www.ps3updates.com/the_2Dgetaway_2D3_2D20050516072440159.jpg
"rendered entirely with Cell", according to Phil Harrison. Now that doesn't look like something a graphically useless CPU could produce, do you agree?
I need a quote to believe that. In every interview I've seen, like
http://www.eurogamer.net/article.php?article_id=59243, he says nothing of the sort, but I can see how people could be mislead into thinking that. He specifically mentioned how a STI terrain demo was done entirely on Cell, but that's it. Even the PR talk about skin shading is pointless, because in 2002 it was found you could
precompute a lot of this using the PRT work from P.P. Sloan et al.
Most of your CPU and GPU philosophies seem to be stuck on current PC ideas. Consoles are certainly not PCs, especially the Playstation consoles. The CPU was the main driving force behind the graphics in the PS2, and it worked out well for them. Why not the PS3? They're certainly touting the Cell as a very graphically capable CPU, afterall, wasn't one of their original ideas to use a Cell as a GPU?
Yes, it was one of their ideas. Guess what? PS3 HAS A NVIDIA GPU IN IT. Why? Cell will never be in the same league as RSX for rendering. Things change, especially in this industry. If you go back to the mid nineties, graphics accelerators were often slower than CPU rendering. Now they're orders of magnitude slower. When Cell was started, NV20 wasn't even released, let alone the revolutionary FP pixel processing of R300. From R300 to G70, rendering speeds have increased 5-fold, but CPU speeds went up by less than 2 times.
You know what? It's worthless to keep arguing with you. Obviously your tangible knowledge of digital hardware and their architecture, the 3D graphics pipeline, and 3D software along with its math is next to zero.