Parallel processing already has proven itself in V3.7, so getting 512 cores to
work parallel should get us impressive results I guess.
It probably would not be possible to let every core process a pixel, but maybe
some pipelining would work were one core processes certain aspects and passes
the result onto a queue that the next core picks up and processes farther, all
running mini algorithms that fit in their own private cache.
see http://www.nvidia.co.uk/page/tesla_computing_solutions.html or
Post a reply to this message