AMBER Archive (2009)

Subject: RE: [AMBER] Is it taking too much time

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Mon Nov 02 2009 - 17:36:48 CST


Hi Nicholus,

I suggest taking a look at the AMBER website which contains a number of
benchmarks for different size systems and different computational setups. If
your machine is a quad core system you should be able to build the parallel
version of sander using MPI. See the manual for instructions. Additionally I
recommend using PMEMD - see $AMBERHOME/src/pmemd/ this is a highly optimized
version of sander which supports a subset of the functionality. On your
machine it will probably run somewhere between 150 and 200% faster. I would
take a look at the benchmarks for JAC (DHFR) which is slightly bigger than
your system. You should be able to obtain equivalent speeds.

http://ambermd.org/

Good luck,
Ross

> -----Original Message-----
> From: amber-bounces_at_ambermd.org [mailto:amber-bounces_at_ambermd.org] On
> Behalf Of nicholus bhattacharjee
> Sent: Monday, November 02, 2009 6:40 AM
> To: AMBER Mailing List
> Subject: Re: [AMBER] Is it taking too much time
>
> Dear Nicolas,
> My simulation is giving more or less similar time
> result.
> With a system 15500 atoms in a single processor system of 2.66 GHz (0.8
> nm
> with PME) in a fedora 9 OS with GNU compiler it is taking 6 hours to
> complete 150 ps simulation. My computer has 4 processors but the job is
> running in 1 processor only. By any means can I distribute by jobs in
> all
> the processors.
>
>
> On Mon, Nov 2, 2009 at 7:49 PM, Nicolas SAPAY
> <nicolas.sapay_at_cermav.cnrs.fr>wrote:
>
> > Hello Nicholus,
> >
> > This is not that easy to answer to your question since the
> performance of
> > your calculation depends on many factors, including:
> > - the number of processors you are using;
> > - their frequency;
> > - the hardware/software used by processors to communicate;
> > - the number of particles in your system;
> > - whether you are using PME or not;
> > - ...
> >
> > I made a test recently with a 30000 atoms system on a single Intel
> proc at
> > 2.4 GHz (cutoff at 0.8 nm with PME). The OS was Fedora 10 and I used
> the
> > GNU compiler. I was able to compute 0.41 ns per day which is not
> > particularly great but sufficient to setup, minimize and equilibrate
> a
> > system (in bulk water).
> >
> > Cheers,
> > Nicolas
> >
> > > On Mon, Nov 2, 2009 at 1:01 PM, manoj singh <mks.amber_at_gmail.com>
> wrote:
> > >
> > >> You can do a quick benchmarking for your system by running few ps
> > >> simulations. The simulation time is proportional to the length of
> the
> > >> simulation.
> > >>
> > >> On Mon, Nov 2, 2009 at 1:08 AM, nicholus bhattacharjee <
> > >> nicholusbhattacharjee_at_gmail.com> wrote:
> > >>
> > >> > Dear community,
> > >> > I am running a protein simulation in my
> IBM
> > >> 3400
> > >> > server (specification
> > >> > http://www-
> 07.ibm.com/systems/in/x/hardware/tower/x3400/specs.html)
> > >> with
> > >> > AMBER 9. My protein is 50 amino acid long. I am doing an
> explicit
> > >> solvent
> > >> > simulation. Can anybody suggest what is the maximum time needed
> for
> > >> this
> > >> > simulation.
> > >> >
> > >> > --
> > >> > Nicholus Bhattacharjee
> > >> > PhD Scholar
> > >> > Department of Chemistry
> > >> > University of Delhi
> > >> > Delhi-110007 (INDIA)
> > >> > Phone: 9873098743(M)
> > >> > _______________________________________________
> > >> > AMBER mailing list
> > >> > AMBER_at_ambermd.org
> > >> > http://lists.ambermd.org/mailman/listinfo/amber
> > >> >
> > >> _______________________________________________
> > >> AMBER mailing list
> > >> AMBER_at_ambermd.org
> > >> http://lists.ambermd.org/mailman/listinfo/amber
> > >>
> > >
> > >
> > > My machine is taking 1/2 hour for 10 ps simulation. Is this very
> long?
> > > --
> > > Nicholus Bhattacharjee
> > > PhD Scholar
> > > Department of Chemistry
> > > University of Delhi
> > > Delhi-110007 (INDIA)
> > > Phone: 9873098743(M)
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER_at_ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> >
> >
> > --
> > [ Nicolas Sapay - Post-Doctoral Fellow ]
> > CERMAV - www.cermav.cnrs.fr
> > BP53, 38041 Grenoble cedex 9, France
> > Phone: +33 (0)4 76 03 76 44/53
> >
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER_at_ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
>
> --
> Nicholus Bhattacharjee
> PhD Scholar
> Department of Chemistry
> University of Delhi
> Delhi-110007 (INDIA)
> Phone: 9873098743(M)
> _______________________________________________
> AMBER mailing list
> AMBER_at_ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber