Dear Dave,
Please pick up the performance patch I prepared for GB calculations on
Altix systems.
You will find the patch in the amber page: http://amber.scripps.edu/
in the section "Amber-related links" under the heading "Efficiency patch
for GB simulations on IA-64, prepared by Roberto Gomperts."
Please let me know if this aleviates the problem you are seeing
somewhat. If not, please contact me directly to see if I can help to
figure out what may be going on.
Thanks,
Roberto
David.Maxwell.di.mdacc.tmc.edu wrote:
> Hi Amber Users,
>
> I noticed something strange when running a small protein GB/SA job on both
> an 4-processor Altix (1.4 Ghz Itanium2) and a 4-processor Tezro (700 Mhz
> MIPS) based SGI. Normally, I get a 2-2.5 speed difference in favor of the
> Altix with running jobs with AMBER 8, submitting with mpirun -np4. This
> makes sense based on the benchmarks I have seen. However, if I run
> another set of jobs of the Altix at the same, the Altix seems to take
> between 3-4 times as long to get done. This is not at all what I would
> expect. I don't seem any memory, disk, or NFS issues. I am running many
> other jobs at the same time, so it is difficult to get an exact figure,
> but obviously something strange is happening. Any ideas on what could be
> causing this?
>
> -Dave
> _____________________________________________
> David Maxwell, Ph.D.
> Senior Research Scientist
> Dept. of Experimental Diagnostic Imaging
> University of Texas M.D. Anderson Cancer Center
> 8012 El Rio St.
> Houston, TX 77054
> Phone: (713)-563-9963 Fax: (713)-563-0084
> david.maxwell.di.mdacc.tmc.edu
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
--
Roberto Gomperts VNET: 483-8851
One Cabot Road, Suite 250 Phone: (978) 562-8851
Hudson, MA 01749 Fax: (978) 562-7450
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Thu Aug 04 2005 - 13:53:00 PDT
