Learning Forum

[akumar44]

Hi Everyone,

I can successfully run the Rocky software on the HPC using two Graphics Processing Units (GPUs). However, when I attempt to use more than two GPUs (three or four), the Rocky software fails to run on the HPC. I am unsure of the cause of this issue.

I have attached the commands used for both the two-GPU and the multi-GPU cases for your reference. Please let me know if you have a solution to this problem.

With this command, I am able to successfully run the Rocky software on the HPC using 2 GPUs:

module load ansys/25.1
Rocky --simulate "/gpfs_common/share03/electrospinning/akumar44/7_30_2025/K_SE_80MJ_CIO.rocky" >"/gpfs_common/share03/electrospinning/akumar44/7_30_2025/log01.txt" --resume=1 --use-gpu=1 --gpu-num=0  --gpu-num=1  --multi-gpu-slicing-direction=X

When I used more than two GPUs and the simulation failed to run, I used the following command:

module load ansys/25.1
Rocky --simulate "/gpfs_common/share03/electrospinning/akumar44/7_30_2025/K_SE_80MJ_CIO.rocky" >"/gpfs_common/share03/electrospinning/akumar44/7_30_2025/log01.txt" --resume=1 --use-gpu=1 --gpu-num=0  --gpu-num=1 --gpu-num=2  --gpu-num =3  --multi-gpu-slicing-direction=X

[Jackson Gomes]

Hi akumar.

The error may be caused due to the syntax error in your multi-GPU command.

You wrote:--gpu-num =3 (with a space before =) maybe this is breaking the parsing and prevents Rocky from starting.

Also ensure your HPC job actually allocates 3–4 GPUs on the same node.

[akumar44]

Hi Jackson,

I appreciate your response.

I used the same node, and there was no space in --gpu-num=3. The space was just a mistake when I wrote the question to you.

Could you please let me know the actual reason why the Rocky software did not run with 3 or 4 GPUs? I was using the 2025R1 version of the Rocky software.