[gmx-users] MPICH or LAM/MPI

[Carsten Kutzner]

hseara at netscape.net wrote:
> 
> Thank you very much you the advise, I will install fiors LAM and then 
> MPICH. By the way you think can be interesting to invest in some low 
> latenci network devices for this little clustter??. Or with the gigabit 
> Can be enought
> 
> Thanks Hector

Of course the scaling will be better with low latency interconnects. If 
you only want to run one large simulation at a time using the whole 
cluster, investing in a good interconnect is a good idea. But if you 
anyway have several simulations running in parallel, it could be better 
to use the money to buy addional computer nodes.

Carsten

> 
> -----Original Message-----
> From: Carsten Kutzner <ckutzne at gwdg.de>
> To: Discussion list for GROMACS users <gmx-users at gromacs.org>
> Sent: Mon, 19 Jun 2006 10:28:51 +0200
> Subject: Re: [gmx-users] MPICH or LAM/MPI
> 
>  Hello Hector,
> 
>  since it does not take long to install lam and mpich, I would install 
> both MPI implementations and then benchmark with a typical MD system 
> which one performs better.
> 
>  I would suggest LAM 7.1.2 which is the newest version, and MPICH-2 
> 1.0.3, which works well, at least on our cluster. Many users (including 
> myself) found that GROMACS on top of MPICH-1.2.x hangs when executed on 
> more than 4 CPUs.
> 
>  My guess is that LAM-GROMACS outperforms MPICH-GROMACS on a smaller 
> number of CPUs, but on 6 or more CPUs MPICH-GROMACS might be faster, 
> since it provides optimized collective communication routines.
> 
>  If you only want to install a single MPI implementation, I would choose 
> LAM for a start.
> 
> Hope that helps,
> Carsten
> 
> hseara at netscape.net wrote:
>  > Dear Gmx-Users,
>  > > I'm configuring a small cluster of 5 Dell Power Edge 1800 with dual 
> xeon > conected by a gigabit dell switch. I was wondering which MPI > 
> distribution, LAM/MPI or MPICH(which version), on this platform has > 
> better performance for gromacs 3.3.1 on a fedora 5 linux distribution. > 
> I will try to combine that with SGEgrid as a queue manager. Also I will 
>  > like to know if I can expect for big systems, 200.000 atoms, good > 
> parallelitzarion in the 5 nodes ( 10 processors).
>  > > Thank you
>  > Hector Martínez-Seara Monné
>  > University of Barcelona
>  > _______________________________________________
>  > gmx-users mailing list gmx-users at gromacs.org
>  > http://www.gromacs.org/mailman/listinfo/gmx-users
>  > Please don't post (un)subscribe requests to the list. Use the www > 
> interface or send it to gmx-users-request at gromacs.org.
>  > Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> 
> -- Dr. Carsten Kutzner
> Max Planck Institute for Biophysical Chemistry
> Theoretical and Computational Biophysics Department
> Am Fassberg 11
> 37077 Goettingen, Germany
> Tel. +49-551-2012313, Fax: +49-551-2012302
> http://www.mpibpc.mpg.de/research/dep/grubmueller/
> http://www.gwdg.de/~ckutzne
> 
> _______________________________________________
> gmx-users mailing list gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
>  Please don't post (un)subscribe requests to the list. Use the www 
> interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> 
>   _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the www 
> interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php

-- 
Dr. Carsten Kutzner
Max Planck Institute for Biophysical Chemistry
Theoretical and Computational Biophysics Department
Am Fassberg 11
37077 Goettingen, Germany
Tel. +49-551-2012313, Fax: +49-551-2012302
http://www.mpibpc.mpg.de/research/dep/grubmueller/
http://www.gwdg.de/~ckutzne