[gmx-users] Varying results due to parallel processing (1 node versus 2 nodes)

[Mark Abraham]

Hi,

This is normal. mdrun trajectories are not reproducible by default, and
reproducing any single trajectory is not generally useful. For example, you
will see similar variation if you run each of those again with different
starting velocities. Any single trajectory is questionable, and you now
have evidence to show that your system hasn't converged this observable on
the 100 ns timescale. You need more of them, or longer simulations,
depending what you are trying to observe.

Mark

On Mon, Sep 19, 2016 at 11:11 AM Sanket Ghawali <sanket.ghawali at gmail.com>
wrote:

> Dear Gromacs users
>
> I am trying to simulate peptide in 30% TFE/water complex. Using parallel
> processing, I ran this system for 100ns using 2 nodes. The peptide is known
> to obtain helical conformation in this system. After 100ns, the peptide was
> found to be 37% helical. I ran the same system again on 1 node instead of
> 2, with no changes in the parameter or the input files. After 100ns of the
> simulation, the peptide was found to be 44% helical. Why is there so much
> change in the results? Which of the two should be considered?
>
> The following commands were used:
>
> Simulation run on 1 node:
>
> mpirun –np 12 –hostfile host /share/apps/gromacs/bin/mdrun_mpi –v –deffnm
> peptide
>
> Simulation run on 2 nodes:
>
> mpirun –np 24 –hostfile host /share/apps/gromacs/bin/mdrun_mpi –v –deffnm
> peptide
>
> Kindly help.
>
> --
> Sanket B Ghawali
> Junior Research Fellow
> ICMR-Biomedical Informatics Centre
> National Institute for Research in Reproductive Health
> Parel, Mumbai.
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.