Try 1 gpu per node. You're clogging the PCIE bus right now... Also AMD
CPUs are the pits.
On Apr 17, 2012 11:26 PM, "Albert" <mailmd2011.gmail.com> wrote:
> Hello:
>
> I am submitting jobs at Forge
> (https://www.xsede.org/web/guest/ncsa-forge) which use GPU and I've made
> some test for a 50,000 atoms protein/water system,
> command:
>
> module load mvapich2-1.8a1p1-open64-4.5.1-cuda-4.1.28
>
> mpirun_rsh -np ${NP} -hostfile ${PBS_NODEFILE}
> /usr/apps/chemistry/Amber/amber11_1.5/bin/pmemd.cuda.MPI -O -i
> prod01.in -p bm.prmtop -c eq2.rst -o prod01.out -r prod01.rst -x
> prod01.mdcrd
>
> here some results:
>
> nodes efficiency (ns/day)
> 1X8 16.44
> 2X8 16.47
> 3X8 16.07
> 4X8 15.17
>
> 1X6 17.98
> 2X6 19.41
> 3X6 20.13
> 4X6 19.70
> 5X6 19.62
> 6X6 19.03
> 10X6 18.33
>
> It seems that the efficiency is not so high and the best one is 3X6 with
> around 20.1 ns/day. Since I am going to run hundreds of ns, it would
> take such a long time to be finished.....
>
> Does anybody got any idea how to improve the efficiency for this CUDA
> running?
>
> thank you very much
> best
> Albert
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Apr 18 2012 - 05:00:04 PDT
