On Tue, Sep 09, 2025, YASHIKA . via AMBER wrote:
>I am running an energy minimization using pmemd.MPI (AMBER20 compiled with
>Intel compiler 18.0.3 and Intel MPI 2018).
>time mpirun -np 32 pmemd.MPI -O -i min_500.in -o min_500.out -c
>with_184.rst7 -p with_184.parm7 -r min_500.rst7 -inf min_500.info -ref
>with_184.rst7 -x min_500.mdcrd
We don't know for sure, but you should use fewer MPI threads. Start with a
serial run, and slowly increase the number of threads (2,4,8,...) to see if
you find an optimal value. Note that generally, pmemd does not scale well
beyond about 16 threads. But the crash problem seems to be related to
minimization.
In general, you (and others) should use sander with ntmin=3 to carry out
initial minimizations. You will almost always get much better results.
...good luck...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Sep 11 2025 - 09:00:02 PDT
