[vivirichard007]

Hi,

I'm testing Converge CFD on cluster with 4 nodes & 2 nodes. I'm experiencing slow progress and less usage of CPU
Model name: Water_pump_rotary_MRF_transient_RANS

Node Configuration: AMD HBv3 64 CPU in each node

************************************************** **********
Slurm run script:

#!/bin/bash
#SBATCH --job-name=ConvCFD
#SBATCH --partition=hpc
#SBATCH --nodes=4
#SBATCH --ntasks-per-node=64

cd /home/ConvergeCFD_Test/Water_pump_rotary_MRF_transient_RANS

MACHINES=$(srun hostname | sort | uniq -c | awk '{print $2 ":" $1}' | paste -s -d "\n" -)

for i in "${MACHINES[@]}"
do echo "$i">hostfile
done

echo $MACHINES

#Run Converge CFD
export RLM_LICENSE="2765@10.32.0.4"
source /home/Convergent_Science/Environment/scripts/CONVERGE/CONVERGE-MPICH/3.1.5.sh
mpirun -hostfile hostfile converge-mpich --super &> logfile_4Node_Water_Pump.out

************************************************** *********
ASK:

I can see all the CPUs are running with Converge CFD, but not 100%, all the CPUs are utilizing only about 30-40% and overall consumption also around 40%.

Is there any way to utilize the full potential of CPU from Convergent Science side???

Note: < dev/null> is not working. If we use this, the simulation is not at all starting.

Kindly do the needful.

[vivirichard007]

HI,

I solved the issue by using Intel MPI instead of MPICH. I have infiniband network and Intel MPI will work better with it.

Thanks.

[ChadTellier]

If we use this, the simulation is not at all starting.

[psrikanth]

Hi,

Are you able to run your case with other mpi versions (mpich?)? What error message do you get?

Praveen

[vivirichard007]

Originally Posted by psrikanth

Hi,

Are you able to run your case with other mpi versions (mpich?)? What error message do you get?

Praveen

Hi,

I'm not getting any error if I use MPICH, just the CPUs are not utilized fully to 100%. When I changed the MPI to IntelMPI, all the CPUs are utilized fully.