posted on Jan, 12 2011 @ 10:55 PM
"Stream" computing is a little different than x86-64 computing. Your graphics card is 'faster' than your CPU - even if you have a six-core
behemoth of a processor. The reason we don't replace your CPU with a graphics card is because graphics processors tend to suck at more complex
processes and one-at-a-time operations (such as user interface... you can only do so much at one time, and will never be able to input the kind of
volume that would benefit from massive parallel processing).
AMD and Nvidia, both, have 'stream' computing solutions. Fire/AMD Stream for AMD, and Tesla (based on CUDA) for Nvidia. It's really not all that
necessary unless you are simulating particle flow or doing some complex engineering simulations.
The main problem facing RISC-based parallel computing today is the lack of programmers for it. Just like there's absolutely no reason -not- to
simply re-compile a 64-bit executable for most games (as many of today's games could benefit from the extra room - I will be pissed if Bethesda makes
TES V 32 bit; Oblivion sprints into the memory limits of 32-bit already and is 4 years old - 64-bit multi-threaded operation is almost required for
any advancement in the game capabilities).
Few programs really need such processing capability and few are willing to jump out there and support it.