So what sort of real numbers you are getting from, let's say, 7950 and 7970? And how it compares to Nvidia cards? If I understand correctly you run multiple WUs at the time to keep it busy but then you have to feed it with CPU cores? I've noticed when tried my GTX580 that the WUs were ready after ~2 minutes and another 20sec it was sitting at 99%. Would running more WUs make it more efficient as they would "overlap"?
Bookmarks