GROMACS version: 2023.2
GROMACS modification: No
I am trying to benchmark GROMACS 2023 on a cluster. I am using the [MPI benchmarks](https://www.mpinat.mpg.de/grubmueller/bench)
I have not changed any parameters in the mdp files. My SLURM submission script assigns 4 GPUs and 72 CPUs per task. This is the command I am using:
gmx mdrun -ntmpi 6 -ntomp 12 -s topol.tpr
This is my error:
Command line:
gmx mdrun -ntmpi 6 -ntomp 10 -s topol.tpr
Back Off! I just backed up md.log to ./#md.log.5#
Reading file topol.tpr, VERSION 2020 (single precision)
Note: file tpx version 119, software tpx version 129
Changing nstlist from 10 to 100, rlist from 1.2 to 1.339
On host gpu3 6 GPUs selected for this run.
Mapping of GPU IDs to the 6 GPU tasks in the 6 ranks on this node:
PP:0,PP:1,PP:2,PP:3,PP:4,PP:5
PP tasks will do (non-perturbed) short-ranged and most bonded interactions on the GPU
PP task will update and constrain coordinates on the GPU
GPU direct communication will be used between MPI ranks.
Using 6 MPI threads
Non-default thread affinity set, disabling internal thread affinity
Using 10 OpenMP threads per tMPI thread
Back Off! I just backed up ener.edr to ./#ener.edr.4#
starting mdrun ‘ConvertedSystem’
-1 steps, infinite ps.
Program: gmx mdrun, version 2023.2-dev-20230712-a3d6917-dirty-unknown
Source file: src/gromacs/ewald/pme_redistribute.cpp (line 305)
MPI rank: 4 (out of 6)
Fatal error:
7 particles communicated to PME rank 4 are more than 2/3 times the cut-off out
of the domain decomposition cell of their charge group in dimension x.
This usually means that your system is not well equilibrated.
Can someone explain why am I getting this error and how to fix it? Is this because of the assigned parameters? I have changed the parameters as well and I am still getting such errors (though the number of particles is different). The node contains GPU=8, CPU=126. These are NVIDIA A100 GPUs.