[gmx-users] General conceptual question about advantage of GPUs

Justin Lemkul jalemkul at vt.edu
Tue Apr 9 17:58:22 CEST 2013


On Tue, Apr 9, 2013 at 11:38 AM, Andrew DeYoung <adeyoung at andrew.cmu.edu>wrote:

> Hi,
>
> For the past 2 years I have been running Gromacs on a standard Linux
> cluster
> (with nodes containing 24 CPUs).  As you know, Gromacs scales excellently
> (and is super efficient), and since the CPUs are Intel Xeon 2.4 GHz
> processors, the simulations run quite fast (I can run 10 ns of ~7000 atoms
> in ~2 days on 24 CPUs).  Can I expect GPUs to be any or much faster than
> these CPUs?
>
>
You can try it out yourself for free and see:

http://www.nvidia.com/object/gpu-test-drive.html?cid=gputestdrive

I would expect better CPU performance, honestly. Have you done benchmarking
on your system to see if fewer processors actually achieve better
performance by eliminating some communication overhead? Using more
processors is not necessarily better.


> There is a rumor in the department that GPUs can give a performance
> increase
> of 10-40 times relative to CPUs, although that group is using another MD
> package.  I am curious whether this performance improvement is typical.  (I
> would not expect it for Gromacs, though, since Gromacs is already super
> fast!)
>
>
Relative speedup depends heavily upon what is being compared. Gromacs has a
very fast baseline, so the relative speedup may not be as high as other
codes that are not as intrinsically fast on CPU :) The GPU performance of
Gromacs, in absolute terms, is very impressive. We run simulations of
~150,000 atoms on a single GPU card (which is somewhat outdated) and can
achieve around 10 ns/day. Newer hardware would undoubtedly perform better.


> If you have time, do you know of any review or opinion papers that might
> discuss the advantages (advantages of performance or otherwise) of using
> GPUs over CPUs?
>
>
Hopefully very soon, Erik's webinar from last week will be posted online (
http://www.gputechconf.com/page/gtc-express-webinar.html). I learned a lot
from it, and I suspect others will as well. There is plenty of literature
about GPU development in MD simulations, but again, be aware of unequal
comparisons and baselines.

-Justin

-- 

========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540)
231-9080http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================



More information about the gromacs.org_gmx-users mailing list