We all know, that ATI is very good on paper in terms of theoretical TFLOPS and also very good in gaming and 3d graphics in general. Still, in folding, ATI lags badly behind NVIDIA. Many folders, using ATI cards, are looking forward to Directx 11 and new drivers, hoping to increase their performance. Possibly itss wishful thinking and NVIDIA is still be strong due to architecturaal differences. Here are few very good threads, analyzing this issue:
Shortly, ATI doesn’t have memory near GPU shaders to perform more complicated tasks (as folding) than graphics processing. NVIDIA has more sophisticated multi-level memory architecture, allowing to solve more complex tasks at GPU shader level.
If You ignore all the flaming here, the final thought is here:
Vaulter98c [H]ard|Gawd, 1.4 Year:
The problem isn’t that ATI GPUs can’t store “enough” data, it’s that they aren’t storing “any” data at all right now since F@H doesn’t use the LDS. And a single step of a single GPU workunit doesn’t require a particularly large amount of data storage, especially not with the small proteins that are currently being used for most of the workunits that are in the wild right now. Each shader unit (set of 4 standard FPUs and one special-function unit) has a 16KB LDS in RV770 and 32KB in RV870, which is more than enough to give a significant performance boost to overall work production speed.
Also very good (illustrated) blogpost about ATI GPGPU issue: