Quote Originally Posted by NKrader View Post
not that you need more bandwidth for gpu "folding"
It does make a difference of about 5%-10% each step between PCie2 from x4 to x8 to x16 so I bet by the time you step all the way down to x1 your GPUGrid output would be very inefficient. The GPUGrid team themselves have stopped using these types of specialty systems and have standardized their lab on regular workstation MOBOs.