I just add my feel about the ps3 "cutting edge" technology and the "ps3 will shine later"
For a lot of people it looks clear that the ps3 is overall a more potent system which in raw power which is true.
But the 360 has its strenght:
UMA
more advanced GPU.
Anyway, I agree on the fact that if one dev team manage to "max out" the ps3 and the other one do the same with the 360, the ps3 is very likely to come on top.
But every time I read that ps3 will get better and better while 360 games should hit a wall in quality soon, I really think that Sony marketting was really clever during the E3 2005.
IMHO, right now with my limited knowledge, I wouldn't have a hard time deciding witch system is put to better use (no matter how many legs ones would think is left with):
It's clearly the ps3!
Look at the presentation from insomnia, naughty boys and the like!
Look R&C, Uncharted, GT prologue and the like.
Ok multipatform games are lagging (but statuquo will be here soon).
But I'm not trying to decide this on average.
And when I read what insomnia is up to... I feel like Ms is in need for "ninjas".
On MS side? Bungy for some reason don't even implement tiling!
Most of the coolest stuffs on the 360 runs on the UE III.
Rare did come with some interesting presentation, particles on the gpu, some use of the tesselators, early use of tiling. But come on where is their "technical jewel" game?
MS hopes too differentiate themselves with games running on UE III engine? Epic is willing to make money on every platform.
Ms provides good tools an "easy going" development environment, but this is not enough they need some dev teams to put theirs hands close to the metal if they want something really brilliant.
Make me think of ome Nao comments, on the games is working the rsx was lagging behing the xenos, he put his hand close to the metal... and guess what brilliant as he is he get the code to run faster on the rsx. When some ask if he also tried his best on xenos, he said so but he also said (if my memory is right) that he had no total acces to load balancing and it sounds to me like the metal was more hide to him than it was while working on RSX.
He also hinted at some clever uses of the tesselator, but didn't develop his idea because of NDA.
And I guess the situation is worse for the xenon, the chip is potentially powerful, tough to use, and from what I red... buggy (cache trashing...).
There is a lot less memory on the xenon than on cell:
1Mo of memory is shared between 6 threads / 3 cores
the latency to this memory is pretty high which must hurts even if SMT can help to hide some latency.
single thread perfs must be nothing to write home.
The only strengh is the vmx128 units, but use them efficiently (write properly vectorised code) doesn't seem that trivial and no mater what you still have to do with 1Mo of slow L2 cache.
Ms provides good profiling tools, but I'm sure that making the most out of it is not that easier than make the most out of the cell.
So far it looks like most devs have stuck to coarse-grained multi threading, and haven't put the simd units to good use. None of the 360 devs is close to trying do what insomnia tries to do with rfom II (I might be wrong... but I have a bad feeling about it).
More on topic.
Some devs have argue that rsx has its limitations in regard to xenos and some others have stated that there work around for those issues (both at gpu level or by using the cpu to help).
Anyway, I've never really understand all that rsx mystic... It's a really potent gpu, that's it no matter if it's really close to a pc part.
And by the way g70 is cleary a PC design it will be a long time before sombody manage to prove otherwise