AMBER Archive (2003)

Subject: Re: AMBER: PMEMD Performance on Beowulf systems

From: Carlos Simmerling (carlos_at_csb.sunysb.edu)
Date: Mon Dec 22 2003 - 10:12:37 CST

has anyone tried this on our xeons? I recall
knowing that 4 jobs ran at once and seemed ok
but I don't know if we actually have timings.
can whoever is using the xeons (Dan? Raphael?)
get a benchmark for

- running a 2 cpu job on 1 node

- running a 2 cpu job on 1 node while another
2 cpu job is running on the same node

- running a 4 cpu job on 1 node

- running a 4 cpu job on 2 nodes
(with nothing else running on them)

carlos

> The other interesting thing I noticed is that HyperThreading gave an
unexpected benefit- if I run PMEMD or AMBER
> with np=4 on a 2 processor (+2 HT processors) machine, I actually get a
mild speedup relative to just 2 cpus. This means, for example, that single
> SMP boxes can run ~1ns sims in reasonable time, so we're not really using
the gigabit interconnect much at all.
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu