AMBER Archive (2002)

Subject: Re: look for benchmark with myrinet vs. 100-BT

From: Margaret Cheung (cheung_at_physics.ucsd.edu)
Date: Fri Aug 02 2002 - 13:20:22 CDT

Dear Dr. Wu,

Please check this website
http://keck2.ucsd.edu/howto/myrinet.html
for brenchmarks.

Sincerely,
Margaret S. Cheung
Biophysics Program
Physics Department 0319
University of California, San Diego
9500 Gilman Drive,
La Jolla, CA 92093-0319
http://www-physics.ucsd.edu/~cheung

On Fri, 2 Aug 2002, Jianhui Wu wrote:

> Dear Amber Users,
>
> Quite some time ago, I asked question about linux cluter vs. Origin 300.
> I would like to thank Drs. P. Burger, I. Makino, C. Klein, O. Fossgaard,
> S. Patchkovskii, E. Leitl, R. G. Brown, Jon Wright etc. for their very
> helpful responses. I am totally convinced that I should go for linux
> cluster when I purchase the new system.
>
> I was told that interconnect is important for MD simulation. I wonder if
> people did some comparison between the Gigabit (such as Myrinet) vs. 100
> Base-T ethernet in term of the benchmark for Sander (such as GB model) and
> the parallel scale ratio. As Myrinet is pricy and it will definitely
> improve the performance, I would like to know how much it could improve
> before I spend 'big buck' on it.
>
> Best wishes,
>
> Jian Hui Wu
> Lady Davis Institute for Medical Research
> Montreal, QC
> Canada
>
>