pmemd can be used on more cores than sander for GB calculations. not just
better scaling, but sander has hard limits of # cores for GB (1 per
residue, I think).
On Fri, Oct 5, 2012 at 7:45 AM, Jason Swails <jason.swails.gmail.com> wrote:
> On Fri, Oct 5, 2012 at 12:01 AM, Sajeewa Pemasinghe <sajeewasp.gmail.com
> >wrote:
>
> > Hi all,
> >
> > According to the amber manual the free energy of solvation has a
> > electrostatic and a non-electrostatic part, the non-electrostatic part of
> > which is calculated as a product of a experimentally obtained
> > proportionality constant times the solvent accessible surface area.
> >
> > So if I set gbsa=0, will the non-electrostatic portion not be calculated?
> >
>
> Correct.
>
>
> > I am specifially asking this because pmemd doesn't support gbsa>0 and
> when
> >
>
> pmemd *does* support gbsa>0 (Amber 12). pmemd.cuda does not.
>
>
> > I use sander, the calculation is so very slow. When I have gbsa=0 and run
> > with pmemd it is very fast but I wanted to make sure how accurate it
> would
> > be.
> >
>
> pmemd and sander use effectively the same GB code on GPUs. I've noticed
> only a marginal speedup using pmemd vs. sander.
>
> If, however, you are referring to the GPU version, then the speedup is
> considerable, as you have pointed out.
>
> HTH,
> 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
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Oct 05 2012 - 05:00:03 PDT
