Re: [AMBER] performance of pmed.cuda.MPI

From: Ross Walker <ross.rosswalker.co.uk>
Date: Fri, 21 Sep 2012 10:22:28 -0700

Dear Jonathan,


>ON a slightly related note:
>
>When running gpu calculations, has anyone looked at or have an idea of the
>effect that the CPU speed has? For example -- if you were to compare
>machines that were identical (same motherboard, hard drive, RAM, and a
>680X) but had different CPU's would you see a performance bump/drop or is
>it still a level playing field? Can you get by with an i3 or i5 instead
>of
>a top end i7? (could be another way to save $ if your building on a
>budget)
>
>any thoughts or insight?

For AMBER the CPU speed will have no impact on single GPU calculations.
You just need at least 1 core per physical GPU. Thus just buy the lowest
bin cheapest CPUs you can. We tend to go with the AMD processors since
they are dirt cheap compared with Intel. I tend to get slightly higher
clock speeds and the 8 core versions of these since that only adds a few
tens of dollars and means the machine has some grunt for running CPU
calculations as well if one wants to do that. If you go with Intel top bin
parts can costs 10x or more lower bin parts sometimes.

For multi-GPU runs (i.e. pmemd.cuda.MPI) then the front side bus speed can
make a difference and so going for the ultra top end chips can help but it
is on the order of 10% or so for a massive increase in price so that
really doesn't help much either.

The spec I have on my shopping list page:

http://www.rosswalker.co.uk/current_amber_gpu_spec.htm

Is pretty much optimized in terms of price / performance for GPU
calculations. Don't expect to run multi GTX680 runs with this setup but it
will scream for 2 independent runs and is not too shabby for CPU runs
either. You can get cheaper by getting a lower spec power supply, 2GB
GTX680 cards etc but then you start cutting into the bone.

Hope this helps.

All the best
Ross

/\
\/
|\oss Walker

---------------------------------------------------------
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
| http://www.rosswalker.co.uk | http://www.wmd-lab.org |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
---------------------------------------------------------

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.






_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Sep 21 2012 - 10:30:03 PDT
Custom Search