Hi there,
I administer a small HPC cluster within my organization and have been trying to get Abaqus to take advantage of CUDA GPUs for analysis for our users.

We have 1 node that is running 2 x AMD Epyc 7502 32 core processors and 2 x Nvidia V100S 32GB GPUs

I have created a custom env file with the following content;

ask_delete=OFF
mp_mpirun_options = "-prot -aff:automatic:bandwidth:socket"
import os
os.environ['ABA_SRM_BALANCED']='ON'
os.environ["ABA_ACCELERATOR_TYPE"]="PLATFORM_CUDA" # Nvidia

And when I run a job with the following command;

abaqus job=[job_name] cpus=[no_of_cpus] gpus=[no_of_gpus] interactive I am seeing the following result in the output;

Begin Analysis Input File Processor
Thu 29 Oct 2020 10:15:14 AM ACDT
Run pre
Thu 29 Oct 2020 10:15:15 AM ACDT
End Analysis Input File Processor
Begin Abaqus/Standard Analysis
Thu 29 Oct 2020 10:15:15 AM ACDT
Run standard
USING ACCELERATOR PLATFORM_CUDA

However when I watch the nvidia-smi on the server itself, I can see that all 32GB of memory for the card is assigned and a process begins, however the GPU utilization is pinned at 0%

I'm not sure if there is some configuration that I'm missing, or potentially there is a licensing constraint preventing me from using GPU but was hoping that someone might be able to assist?

Cheers,
Ryan