Gone are the days when GPUs were only used to make your favourite games look great, using GPUs to do actual computing tasks is the in thing. GPUs being harnessed for number crunching have already helped push the folding@home project past 4 Petaflops, the GPU client alone counts for around 2 Petaflops, for comparison that’s more than twice the computing power of the worlds fastest supercomputers.
Sticking with the folding@home client for now, currently there are around 20,000 GPU clients ‘folding’ producing around 2 Petaflops (2000 Teraflops) and we’ll assume for arguments sake that the average folder’s PC cost around £500, which is a little high as one can make a system capable of running the client for quite a bit less and I’m not counting montiors etc, but we’ll assume and average gamer. 20,000 x £500 = £10,000,000. Each Blue Gene/L (one of the top supercomputers around) hits around 6 Teraflops per rack each rack consisting of 1,024 dual core PowerPC based processors and cost (2004 data) around £1,000,000 each. Matching the price of the GPU folders Blue Gene/L racks thats only 10 racks pumping out 60 Teraflops, further; to reach 60 Teraflops with GPU folding the cost is around £300,000.
More to come on this, some of us have to sleep.