Hi all,
I've got a renderfarm setup running Centos 5, stock kernel with 4 core x 2 cpu Xeon E5450 (total 8 cores) 8GB RAM. The version of Maya is 8.5sp1
I used the command:
time Render -renderer sw -s 101 -e 101 -proj /project -rd /tmp testscene.mb
to render.
The output of time is as follows [edited 20081302: pasted the wrong times, corrected]:
real 6m29.232s
user 15m49.178s
sys 24m59.942s
Just as a comparison, I ran the same command on a dual core dual cpu AMD (4 cores total)box w/ 4G RAM [edited 20081302: pasted the wrong times, corrected]:
real 7m0.631s
user 22m6.719s
sys 2m7.900s
But when I ran the same command twice (in parallel, by backgrounding one process and starting another right away) on the 8 core box, I get the following for each of the maya process:
real 5m15.583s
user 13m16.546s
sys 4m52.708s
real 5m15.583s
user 13m16.546s
sys 4m52.708s
So I did a strace -Ttt -f to see what's going on and analyzed the output.
CALL cpu(sec) percent
futex 122.12 84.03%
read 5.19 3.57%
mmap 4.63 3.19%
brk 3.76 2.58%
My issues are that first of all, there doesn't seem to be much performance gained between 4 cores and 8 cores, and secondly why is it that running 2 processes of maya in parallel result in faster rendering time? Bios setting? Kernel configuration options? Anyone have any ideas?