AMBER Archive (2009)

Subject: RE: [AMBER] Computational Resources

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Wed Aug 05 2009 - 11:08:21 CDT


Hi Bill,

> One time, I read AMBER has a limit for computational resources, it
> means that there is no a big different in time when you run your job on
> 64 processors and 128 processors (FOR EXAMPLE). Is that right? and if
> yes, what is the maximum number of processors should I use to save our
> computational resources and don't waste time?

This is a very complicated question that depends on a huge number of factors, not least the specific problem size and type of run you are doing, whether you are using sander or pmemd, what your hardware is, what the interconnect is etc etc.

In short if you can you should be using PMEMD. This will perform much better than sander and will scale a lot further. If you are running replica exchange then generally you can run a lot more processors (in sander) than you would for a standard MD run. Your best bet is to take the system you want to run on the machine you plan to use a run some short benchmarks to check performance and scaling.

For reference you can take a look at:

http://ambermd.org/amber10.bench1.html

For the U.S. TeraGrid supercomputers see:

http://www.teragrid.org/userinfo/aus/amber_benchmark.php

These timings are all for standard (PME) MD runs using PMEMD v10.0.

All the best
Ross

/\
\/
|\oss Walker

| Assistant Research Professor |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
| http://www.rosswalker.co.uk | PGP Key available on request |

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_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber