Dear Ambers,
Thank you for kind support everytime.
I have one question for unbalanced usage of GPU cards in Replica Exchange MD in Amber22.
Although I set two GPU cards for pmemd.cuda.MPI to run REMD with 50 replicas,
I found that one GPU card has larger number of processes than the other
(50 v.s 25).
and the former GPU memory was twice consumed in the simulation
( 20597MiB / 24576MiB v.s 12839MiB / 24576MiB).
I'm supposing if I can use the two GPU card equally for pmemd.cuda.MPI
I could employ greater number of replicas for REMD in my simulation platform.
Is that possible to solve such an unbaance with some setting in Linux, or Amber option?
I'm most grateful if you give an advice for this problem.
Yours sincerely,
Kurisaki
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Nov 08 2022 - 00:00:03 PST
