[gmx-users] gromacs cpu v/s gpu

Tsjerk Wassenaar tsjerkw at gmail.com
Wed Jan 15 10:17:27 CET 2014

Hi Gurunath,

You have to run both simulations for (approximately) infinite time and
compare the probability density functions to check whether the simulations
are the same. Maybe running more simulations would also be a good idea.
Here's something about simulations and statistics:
http://onlinelibrary.wiley.com/doi/10.1002/jcc.20341/full(sel-advertisement ;))



On Wed, Jan 15, 2014 at 10:05 AM, Gurunath Katagi <gurunath.katagi at gmail.com
> wrote:

> Hello everyone...
> I have done a MD simulation of peptide using gromacs 4.6.5 both on CPU
> based system (intel based)  I have run the same simulation in the GPU
> based system (NVIDIA quadra ) to test whether i get the siulialr results. I
> have taken care into the necessary modifications in the mdp files lile
> cut-off scheme.
> But the results from the 50ns all-atom simulations with OPLS forcefield
> seems to be different . i.e the conformations that are resulting from the
> two runs are different...
> Can anybody suggest what other things are to be taken care apart from mdp
> file modifications while running on gpu, because the cpu based simulation
> run is giving the expected results.
>  So now i need to utilize the gpu based system for further computing, i am
> just wondering how can reproduce the results in gpu based system...?
> Thank you
> Gurunath
> --
> 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.

Tsjerk A. Wassenaar, Ph.D.

More information about the gromacs.org_gmx-users mailing list