Re: [AMBER-Developers] About cutoffs in sander

From: David Cerutti <dscerutti.gmail.com>
Date: Mon, 13 Mar 2017 22:57:36 -0400

pmemd CPU can deal with multiple cutoffs, and this is a way to improve
performance in some situations (low numbers of cores, EE < LJ cutoff), but
pmemd GPU cannot. You can go up to 20% faster in those special cases by
cutting EE down to about 6.5, but you have to be careful--there has to be a
tradeoff in reciprocal space work or you get burned, so order = 6 in &ewald
when you go EE cutoff = 6.5. Old school tricks, though...

Good luck,

Dave


On Mon, Mar 13, 2017 at 10:42 PM, pengfei li <ldsoar1990.gmail.com> wrote:

> Hi Jason,
>
> Thanks for your reply. I am going to do QM/MM calculations, since CHARMM
> was used in our group before, I want to use AMBER to do the calculations
> but use similar conditions. It seems like the CHARMM protocol in our group
> uses different cutoffs for electrostatics and VDW so I am wondering whether
> AMBER also has this option available.
>
> Kind regards,
> Pengfei
>
> > On Mar 13, 2017, at 9:30 PM, Jason Swails <jason.swails.gmail.com>
> wrote:
> >
> > On Mon, Mar 13, 2017 at 1:06 PM, pengfei li <ldsoar1990.gmail.com>
> wrote:
> >
> >> Hi Everyone,
> >>
> >> Can sander deal with different cutoffs for electrostatic and VDW
> >> interactions? Thanks!
> >>
> >
> > ​No. But pmemd can, I think.
> >
> > What's the use-case here? If you use PME, the cutoff doesn't have an
> > appreciable effect on the electrostatics -- it just affects the split
> > between the direct and reciprocal space contributions. So you really
> only
> > need to consider L-J accuracy when picking a nonbonded cutoff. You can
> > test this with pmemd, but I don't think you gain much efficiency by using
> > different cutoffs for L-J vs. electrostatics...
> >
> > Good luck,
> > Jason
> >
> > --
> > Jason M. Swails
> > _______________________________________________
> > AMBER-Developers mailing list
> > AMBER-Developers.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber-developers
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Mar 13 2017 - 20:00:04 PDT
Custom Search