Osx lion & cuda

[Joseph Owens]

Just ran the CUdaZ performance tests on my Q4000 system:

Host-to-device 5808 MiB/s, so not significantly different, however,

Single-precision Float processing is around 482 GFlop/sec, and Double is less than half of the Q4000’s own expectations at around 188 GFlop/sec, compared to the GTX480 (flashed) GPU statistics of 1338, and I see I must have mis-typed the Double precision rate for it, as 168 is not right, it was still right about half. (probably some kind of keyboard dyslexia and its probably more like 681…)

So, just using these raw numbers, flashed-480 appears to be about 3x the performance, but I can’t comment on how that translates wrt Resolve. Still, simply by the throttle numbers, the Q4000 PNY-Apple version appears to be WAY behind the flashed 480, (1/3 the Gflops) but 3 times the price.

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Murray]

Jacob, what CUDA card are you using and does it need both pcie power supplies? If so how did you hook it up with a 5770 as that needs one 6 pin power?