Re: [AMBER] System blows up after several nanoseconds using Amber 12

From: Sasha Perkins <slperks49.gmail.com>
Date: Sat, 10 Nov 2012 12:02:17 -0500

Yes, so we currently have Amber 12 only installed on the GPU clusters and
we are not allowed to run anything on there that does not require GPUs or
the job will automatically get killed. But I'll see if I can get that
exempt so I can try to isolate the problem.

Also, just so you know, when I type in ./patch_amber.py --patch-level on
our amber12 home, the output is:
Latest patch applied to AmberTools12: 26

Nothing comes up regarding amber 12 bugfixes.

I'll also try pmemd.cuda_SPFP for only a single GPU and see if the problem
persists otherwise I may need to only use amber 11 or 10 for this system
(which is 11400 atoms, 48 Angstrom size box).

Thanks for your help!

Sasha


On Sat, Nov 10, 2012 at 11:13 AM, Jason Swails <jason.swails.gmail.com>wrote:

>
>
> On Nov 10, 2012, at 10:32 AM, Sasha Perkins <slperks49.gmail.com> wrote:
>
> > Jason,
> >
> > Thanks for the prompt reply. I agree with you that the cut off is very
> > large. The only reason I'm using that cut off is because in the
> literature,
> > that is the typical cut off for my type of system and so I tried that
> > first.
> >
> > The latest bugfix for Amber12 we have is 26.
>
> Amber 12 does not have that many bug fixes. What is the output of
>
> ./patch_amber.py --patch-level
>
> ?
> >
> > I am using the Penn State Lion-GA cluster with only one GPU for now
> > using pmemd.cuda_SPFP.MPI. The Lion-GA cluster GPUs I believe are Nvidia
> > Tesla M2090. Should I be using a different command if I'm only running in
> > one GPU? Or simply use more GPUs?
>
> Use pmemd.cuda_SPFP for only a single GPU.
>
> As Adrian said, though, try the CPU version of Amber12 first to isolate
> the problem to GPUs.
>
> Good luck,
> Jason
>
> --
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Candidate
> 352-392-4032
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



-- 
Sincerely,
Sasha Perkins
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Nov 10 2012 - 09:30:02 PST
Custom Search