[gmx-users] Extension on different machines (CPU/GPU

[Nicolas Cheron]

Hi Mark,

Thank you for the quick answer. However, the problem is not only a problem
of reproducibility of a single simulation. To make sure that I have
converged results with proper sampling, I am performing each time 10
simulations of the same system with different positions of the solvent
molecules and different initial velocities. For the 10 simulations, I had
reached a plateau at 250ns for the parameters I am looking at. When
extending to the GPU machine, for the 10 simulations I see these parameters
increasing. For other sets of simulations with the same solute but
different water:glycerol molar fractions, the plateau stays stable until
300ns and I never saw this kind of increase. That's why I was wondering if
there is a big red flag when extending simulations on another machine or
switching from CPU to GPU.

Thank you.

Nicolas


Le mar. 5 juin 2018 à 10:43, Mark Abraham <mark.j.abraham at gmail.com> a
écrit :

> Hi,
>
> You'll basically never get a reproducible trajectory
> http://www.gromacs.org/Documentation/Terminology/Reproducibility. If your
> conclusions would depend on the configurations in any single trajectory,
> you'll have a hard time demonstrating that those conclusions are valid
> reflection of reality.
>
> Mark
>
> On Tue, Jun 5, 2018 at 10:37 AM Nicolas Cheron <
> nicolas.cheron.boulot at gmail.com> wrote:
>
> > Dear all,
> >
> > I ran a 250ns simulation with Gromacs 5.1.2 on a machine with only CPUs.
> I
> > then wanted to extend the simulation to 300ns, and copied the .tpr and
> the
> > .cpt to another machine with GPUs (also running Gromacs 5.1.2). I
> observed
> > strange results, so I decided to also extend the simulation (from 250ns)
> on
> > the first machine to compare.
> >
> > Different results are obtained for the 2 extensions (that were started
> from
> > the same .tpr and .cpt) on the 2 different machines/architectures. Is it
> a
> > known behaviour, and one should never proceed like this? Is it due to
> > CPU/GPU, or only the fact that machines are different and the same would
> > have happened on another CPU machine?
> >
> > Thank you
> >
> > Nicolas
> >
> > PS: I am looking at a solute in a water:glycerol mixture. By "different
> > results" I mean that the number of water (or glycerol) around the solute
> > are different at all time steps
> > --
> > 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.
> >
> --
> 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.
>