GROMACS version: 2020.4
GROMACS modification: No
I am running md simulations using gmx on Expanse supercomputer at SDSC. My system has 648, 379 atoms. Running the simulation, on one compute node, returns a performance output of
Core t (s) Wall t (s) (%) Time: 284153.583 35519.199 800.0 9h51:59
(ns/day) (hour/ns) Performance: -18.071 -1.328
I was testing out various configurations (one node, 2 nodes, 1 one node gpu, etc.). Using one gpu provides around 3 ns/day. I am not sure what that negative number means. Is gmx complaining about improper usage of resources, or is something else wrong with the simulation set up?
Any advice is appreciated. If anyone has used the Expanse supercomputer and has some insight into setting up jobs properly, that is also highly appreciated.