[AMBER] different performance with and without distance restraints in TI calculation

From: Qinghua Liao <scorpio.liao.gmail.com>
Date: Mon, 29 Nov 2021 14:07:36 +0100


I am trying to simulate the transformation from one ion to another with
free energy perturbation
using pmemd.cuda of Amber 20 (GPU card V100), to prevent the ions from
getting out of the active sites, some distance
restraints were applied (setting nmropt=1), but the performance is just
20 ns/day. If I just turn of the distance restraints
(nmropt=0), then the performance goes up to 80 ns/day, which is the same
performance of simulating a mutation of
the same system without distance restraints I did some days ago.

Is the performance drop reasonable?

I will also try to use the PMD module (&pmd) to apply the distances.
 From my previous experience, performance drops more
with &pmd than with nmropt=1 for normal MD simulations.

Thanks a lot!

All the best,

AMBER mailing list
Received on Mon Nov 29 2021 - 05:30:02 PST
Custom Search