Quote Originally Posted by dan View Post
Did you look at the reported GPU time for your calculations? I was looking at wall clock time versus GPU time, which is more relevant for an estimate of daily credit. The reported GPU time was usually around 850 to 900 seconds, and wall clock was 1260. I'm running with dual cores and it's just not efficient enough, when compared to Milkyway and wall clock times.

Dan

One other note: Running Moo in CF by itself, I was seeing one full core (25% CPU on an i7) being used to run Moo.
With regards to CPU, I reserve 1 full core (not thread) for the GPU.

Yes, my calculations are based on "wall time" not reported time. On the single 5870 the project actually reports wall time accurately. On dual cards or more what is shown is NOT wall time so you have to look inside the stderr report to get the actual time.

My 5970 (also w/ 4 wu done) projects to 565k/day running at 900 MHz versus 440k on MW.

The only machine that doesn't do better on Moo than MW is my 5970/5870 computer. That is because the wu with odd numbers of packets lets 2 GPUs sit idle while the 3rd one finishes the packet. The BEST configuration by far is a single GPU!