Re: [AMBER] AMBER 11 on multiple GPUs

From: Gould, Ian R <i.gould.imperial.ac.uk>
Date: Wed, 1 Dec 2010 18:04:34 +0000

Hi Ross and fellow GPU AMBER users

I have been using the parallel version for a while and can report that it does exactly what is says on the "Tin".
I only have 2 C2050's in my machine and so I am seeing speed up between 1.5 and 1.66 depending on size of system and cutoff used.
It's worth noting that in general on X64 based machines you rarely see better than 1.7x going from 1 to 2 processors so 1.5 to 1.66 is very good when you think about the power per card that the 2050's represent. I am keen to extend the experiment up to 4 cards. I should also mention that this is all done over PCI-X and not via infiniband. The motherboard I have should enable me to run 4 x C2050's at full 16x on the PCI-X so when I can get hold of another couple of cards I will let you know what I find. I really do hope that NVIDIA sort out the GTX 480/580 issue.

Cheers
Ian

"Education: that which reveals to the wise, and conceals from the stupid, the vast limits of their knowledge."
Mark Twain 
--
Dr Ian R Gould
Senior Lecturer Biological and Biophysical Chemistry
Imperial College London
Exhibition Road
London
SW7 2AY
E-mail i.gould.imperial.ac.uk
http://www3.imperial.ac.uk/people/i.gould
Tel +44 (0)207 594 5809
Fax +44 (0)207 594 5809
On 01/12/2010 17:51, "Ross Walker" <ross.rosswalker.co.uk> wrote:
Dear AMBER users,
I just wanted to let you all know that we have now released a patch as part
of the bugfixes (bugfix.9 and bugfix.11) that provide support for
acceleration of PMEMD v11 on multiple GPUs using MPI. For those that are
interested there are compilation instructions and updated benchmarks on the
following page:
http://ambermd.org/gpus/
This represents the first release of multiple GPU support. We are currently
working on improving performance and scalability further and anticipate
releasing additional updates in a few months.
I would appreciate any feedback.
Additionally we are still working with NVIDIA to address the problem
involving runs hanging on GTX4XX and GTX580 cards and I am hopeful that a
workaround can be found.
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
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Dec 01 2010 - 10:30:03 PST
Custom Search