oh so now Xenon has a TERAFLOP processor? (just CPU alone??)

[blakjedi]

I want a PS3... not because I like Playstation games per se but but I am a "tech whore." I will be buy the Xenon the day it comes out though because I am interested in teh games on that system...

This generation I didnt buy a PS2 because I wasn't interested in either the technology or the games on that system.

 

[sunscar]

I never said it was a count of EVERY operation in the GPU, I said fixed point and floating point, as did Jaws (who I was replying to, btw). Nvidia, I said in the past counted fixed point and floating point math together when giving their performance figures. The trillion opperations per second of the Geforce 4 are largely undefined, and do not represent those two only. It can be taken to mean floating point, fixed point, instructions read, etc. In that regard if Nvidia wanted to call it a trillion ops hardware they would probably be correct, but they'd be damned brazen to call it a trillion floating point ops hardware because it would be wholely incorrect and false. They don't do that, by the way.

So, no, I did not miss it. The geforce 4 is not a trillion flops device and in no way could be labeled as one. Just as the current Geforce 6 series aren't tflops devices and can't be marketed as such. They could be labeled "Tops" devices, but that's not the issue here.

Enough dragging this thread off topic. The entirety of this post is in response to the possibility of the CPU or even the GPU in the next Xbox being a teraflops part, and it simply will not be. Could the entire machine add up to that? Maybe. Will the next playstation CPU or GPU be? Again, probably no. Maybe the whole machine but not even scratching the surface of "likely" on just the GPU or the CPU.

Lastly here, is this thread even relevant anymore? This was discussion over weather or not the speaker (that this entire thing originated over) was misquoted, or misunderstanding, or misinformed. I think the answer to that question has been answered over and over again as "yes". Somewhere along the chain something got terribly fried and the report bacame irrelevant.



Later

 

[randycat99]

of course we all know that a small lowend supercomputer with a 1 TFLOP sustained performance rating, or even a 1 TFLOP peak perfrmance rating, would eat Xenon and PS3 for lunch when running most applications

I dunno- would it??? It's not like these industrial supercomputers use magic processors. They deliver the performance they do because of brute multiplicity of rather average processors and extremely specialized OS+software to exploit scaleability. So running "most applications" would imply "typical applications", hence you would not see much better performance than a "single processor" scenario (not so ironically the same situation you have with these next gen, multiprocessing consoles). Essentially, what we are seeing is a migration of these exotic supercomputer techniques into the consumer realm. Hence the contexts of these "home supercomputer FLOP ratings" will be largely the same as these industrial supercomputer that we have known about for some time.

AFAIK (and I could certainly be wrong on this), the last major, single processor solution (that was intended to deliver a big floating-point wallop with just a single core on a single chip) was Intel's Itanium. It was, indeed, a FLOP monster at mere sub-Ghz clockspeeds (that could utterly annihilate any desktop CPU of its time). The caveat- it was based around a totally new 64-bit ISA that was utterly incompatible with the existing x86 ISA. So it suffered an ironic fate of irrelevance at the feet of an utterly entrenched platform (32-bit x86) which its own creator (Intel) had proliferated with such great success. This is also not to mention the large die and excessive heat issues that plagued it in its era of introduction. I think they are still around, though, implemented on current manufacturing processes. However, the performance crown is stolen once again as the industry moves in the brute multiplicity direction. I'm not really up on what FLOP ratings they are up to, but I don't think they are quite keeping up with the TFLOP on a chip race... If you want a pragmatic view on the Itanium, they seem to be living up (no pun intended) to that evolutionary dead-end, dinosaur-of-a-processor profile.