V-Ray’s Scalability Potential

[ddixon]

I had a question about V-Ray’s scalability potential. Right now we are using V-Ray on a PipelineFX Qube render farm and would like to better facilitate the sharing of computer power on our render nodes by limiting the maximum thread count of our renders to a lower number of threads on a multi core system. For example, limiting a sixteen-core system to four cores to allow for different render jobs to process on the same system. The problem is that though we expected to see a decline in rendering performance do to the reduced number of cores available to a respective render job, the decrease in performance has been substantial per job, even when no other render jobs are currently running on the same system. Is there someway to achieve better render performance from V-Ray using this method?

[vlado]

How are you limiting the threads? The best way would be to change the process affinity to just some CPUs of the system. If you simply limit the number of threads, and your machines support hyperthreading, it may turn out that you are using less physical cores. F.e. if you specify the render to use 4 cores, they might end up being only 2 physical cores (but four logical).

Best regards,
Vlado