Here 46.97 seconds, having a GTX460 1GB, actually I’m using only 50% on my “horsepower” cause I have installed another GTX460 so once Cycles support multi-gpu should be twice faster aprox.
2:40:57
CPU, OSX10.6.7 blender 64bit, 2 nehalem xeon
The superiority of 16theads xeons is obvious, even running at 2.26GHz.
i7 are not xeons I mean.
I love this 46 sec! And the 01:08 with 550ti. The GPU build using a gt120 … lets forget it LOL
Be aware though, there is massive performance difference between builds ( at least on windows ). The ‘cpu_opt’ build from graphicall seems to be the fastest i’ve tried so far, rendering my test scene in literally half the time of one I compiled myself from trunk.
Have been trying to build a version for SSE4 with profile guided optimizations - should be possible, but is proving very tricky :(. If it works, that should be blazingly fast for CPU only
I use multiple cards for a 4 monitor setup, interestingly the Nvidia Control Panel lets you choose what video cards are available for CUDA use, but doesn’t work. I had to disable the GTX285 in device manager to benchmark the 9600GT