[gmx-users] Gromacs slow for 23000 atom DPPC bilayer on a (1 x 4) node: 50 ps in 10 hours

[maria goranovic]

An update:
Using a 4th order PME spline resulted in ~ 15 % increase in efficiency.

Thank you, Carsten.

On Mon, Mar 24, 2008 at 3:08 PM, maria goranovic <mariagoranovic at gmail.com>
wrote:

> Dear All,
>
> My apologies. I had too big a simulation cell, and too few atoms, hence
> the problem.
>
> No particular reason to choose order 5. I will try with pme_order 4 and
> see if it improves performance anyway.
>
> thanks !
>
> -maria
>
>
> On Mon, Mar 24, 2008 at 2:19 PM, Carsten Kutzner <ckutzne at gwdg.de> wrote:
>
> > Am 24.03.2008 um 10:17 schrieb maria goranovic:
> >
> > > Hi Folks,
> > >
> > > My simulation is running too slow. It took 10 wall clock hours (40
> > > cpu hours) for a short 50 ps simulation of a ~ 23000 atom DPPC
> > > bilayer. The hardware is a 4-cpu core. The installation is gromacs
> > > 3.3.1. I have run much larger systems (~ 160000 atoms) using the
> > > same gromacs installation on the same hardware, and they run much
> > > faster than this (200 ps per 40 cpu hours).
> > >
> > > Can anybody suggest why this is happening ? Is it because of latency
> > > in the cpu communication? If so, what is the workaround ?
> > Is there a special reason for using pme_order=5? I would use the
> > default, 4 instead, or at least an even number.
> >
> > Carsten
> >
> > > My .mdp script is below.
> > > These are the run commands.
> > >
> > > grompp -np 4 -v -f heat.mdp -c minim4.gro -p dppc.top -n dppc.ndx -o
> > > heat.tpr
> > > mpirun -np 4 ~/bin/mdrun_mpi -np 4 -v -s heat.tpr -o heat.trr -c
> > > heat.gro -e heat -g heat.log > & heat.out
> > >
> > > ;###########################################
> > > ; heat.mdp
> > > ;
> > > title               =  heating
> > > cpp                 =  /usr/bin/cpp
> > > constraints         =  hbonds
> > > constraint_algorithm = lincs
> > > unconstrained_start =  yes
> > > integrator          =  md
> > > nsteps              =  25000
> > > dt                  =  0.002
> > > comm_mode           =  linear
> > > nstxout             =  5000
> > > nstvout             =  5000
> > > nstlog              =  5000
> > > nstenergy           =  5000
> > > nstlist             =  10
> > > ns_type             =  grid
> > > pbc                 =  xyz
> > > ; ------------------
> > > coulombtype         = PME
> > > rcoulomb            = 1.0
> > > vdwtype             = cut-off
> > > rlist               = 1.0
> > > rvdw                = 1.0
> > > fourierspacing      = 0.1
> > > pme_order           = 5
> > > ewald_rtol          = 1e-5
> > > ; ---------------------------
> > > ; Berendsen temperature and preasure coupling
> > > Tcoupl              =  berendsen
> > > tc-grps             =  DPPC  SOL
> > > tau_t               =  0.6      0.6
> > > ; i have also tried a tau value of 0.1, but no speed up
> > > ref_t               =  323.0     323.0
> > > Pcoupl              =  berendsen
> > > Pcoupltype          =  semiisotropic
> > > tau_p               =  1.0        1.0
> > > compressibility     =  4.5e-5  4.5e-5
> > > ref_p               =  1.0        1.0
> > > ; ---------------------------
> > > gen_vel             =  yes
> > > gen_temp            =  323.0
> > > gen_seed            =  194040
> > > ;###########################################
> > >
> > > --
> > > Maria G.
> > > Technical University of Denmark
> > > Copenhagen _______________________________________________
> > > gmx-users mailing list    gmx-users at gromacs.org
> > > http://www.gromacs.org/mailman/listinfo/gmx-users
> > > Please search the archive at http://www.gromacs.org/search before
> > > posting!
> > > 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 search the archive at http://www.gromacs.org/search before
> > posting!
> > 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
> >
>
>
>
> --
> Maria G.
> Technical University of Denmark
> Copenhagen
>



-- 
Maria G.
Technical University of Denmark
Copenhagen
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20080325/3a2e4552/attachment.html>