I was about to suggest a "feature" for boinc related to statistics. I saw the CPU stats and wondered if they were "skewed" by systems with GPU's or not and wondered why they didn't have GPU stats and then I found them. I don't remember where now but there was a listing of video cards and their "relative" results where NVIDIA and ATI were broken out separately and the values were normalized (ie: the top NVidia card was 1.00 and all other Nvidia card models followed at .98, .90, .88 etc... and then the list started over again with ATI cards with the top card at 1.00 and the rest following at .98, .9x, .8x, ... and so on... Why doesn't the GPU list all cards and their actual ratings so one can see which cards run BOINC the fastest?

Simply said because it depends on which project you run and how well they have optimized their various GPU applications for the various GPUs out there, be it AMD, Nvidia or these days Intel.

So if you want to know well a certain GPU works at a certain project, you best ask at that project, or look on their forums if someone else asked this already.

---

Well, I asked several times in the past and no one could give a definitive answer, at least this list is a step in the right direction. I'm sure there are plenty of each card in use and enough stats to give a hint as to what card perform the best for each project or in general best overall. I was very interested to see if the best ATI and NVIDIA card on the list we around the same place or was one better then the other? Is there a correlation between the listed GFLOPS and WU performance? I have seen that the newest card isn't necessarily the best. For example in the NVIDIA line, the middle digit of the card model is very important. Even though the left most digit is higher for better cards a 690 would out perform a 760. On NVidia is the number of CUDA cores a major factor? (I'm only considering the cards as they come and not extra overclocking).