If performance is paramount, and your systems are not using continuum solvent
(GB), I strongly recommend that you explore PMEMD:
http://amber.scripps.edu/pmemd-get.html
We get a 25% speedup over sander7 on our dual Xeon systems.
Recently published enhancements to PMEMD, as I understand, will enhance it's
scalability to large clusters.
Part of the speedup comes through code streamlining in Fortran 90 coupled with
Intel compiler optimizations.
If you have trouble getting it to run, post to the list, and I'm sure the
author will help out.
Chris
>===== Original Message From "Malone, Scott" <Scott.Malone.stjude.org> =====
>Greeting,
> I'm new to running amber and I'm look for some input from people on
compilers and compiler options that people find to give the performance for
amber7. We are currently running amber on Linux cluster of IBM BladeServers,
each with 2.5 Gb of
>Memory, 2.4 Ghz Zeon Processors, and GigE interconnect. Each blade is
currently running RedHat 7.3 with Kernel 2.4.18-27.7.xsmp. We have both
gnu3.1 and Porland 4.1 compilers (we have pgi5.0, but not installed yet.). Of
course recommendation on
>other compilers would helpful as well..
>
>Also, any thoughts on amber8 and compilers would also be helpful....
>
>Your thoughts on this would be of great help and thanks in advance!
>
>Scott Malone
>Systems Integration Engineer
>Hartwell Center for Bioinformatics and Biotechnology
>scott.malone.stjude.org
>901.495.5957
>
>
>
>-----------------------------------------------------------------------
>The AMBER Mail Reflector
>To post, send mail to amber.scripps.edu
>To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Chris Moth
chris.moth.vanderbilt.edu
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Jan 14 2004 - 15:53:05 PST
