Re: [AMBER] Dynamic at high temperature

From: Carlos Simmerling <carlos.simmerling.gmail.com>
Date: Fri, 25 Nov 2016 13:24:13 -0500

To add to what Dave said, make you you are very careful with high T.
Previous discussions in the archive give more detail. At very high T (above
500 or so) you will sample cis/trans isomerization, chirality inversion and
other things that you may not want. You will need to do very careful and
extensive analysis, and perhaps include many additional restraints during
the simulations. You also may want to heat slowly and don't just assign the
high thermostat as temp0. You also want a structure that is well minimized
and equilibrated prior to the heating.

On Nov 25, 2016 8:44 AM, "David A Case" <david.case.rutgers.edu> wrote:

> On Fri, Nov 25, 2016, Imma Speciale wrote:
> >
> > I have a problem about the running of the dynamics! I want to visualize
> > what happens when my molecule is at high temperature, thus I have put in
> > the first script md1.in tempi=0, temp0=300, in md2.in tempi=300,
> temp0=300
> > and in the md3.in (the real dynamic) tempi=300, temp0=500. In this way
> the
> > dynamic run without problems. But when I change the final temperature
> with
> > 700 or 800 K, the dynamic run and after 900/1000 ps it crashes!! In
> > terminal I receive this message:
> >
> > cudaMemcpy GpuBuffer::Download failed unspecified launch failure
>
> This is a generic failure message. Remember that you may need short time
> steps to integrate systems at high temperatures.
>
> A good debugging strategy is to run the simulation on a cpu: you will often
> get better error messages. It's also possible that you are seeing
> fixed-point
> overflow on the GPU. Running on a cpu may help identify the problem.
>
> It's also unclear whether you have a periodic or a non-periodic system, and
> whether you are running constant pressure simulations or not.
>
> ...good luck....dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Nov 25 2016 - 10:30:02 PST
Custom Search