Starting today, PBS now handles NVIDIA Volta V100 GPU nodes differently than in the past. The new configuration allows multiple PBS jobs to share a single node. For detailed information, see Changes to PBS Job Requests for V100 GPU Resources.

Previous Notification

To enable more efficient use of the NVIDIA Volta V100 GPU nodes on Pleiades, we will reconfigure the nodes to allow multiple PBS jobs to share a single node on Thursday, July 16 at 8:00 a.m., Pacific Time.

Note: These changes do not apply to a few V100 GPU nodes reserved for special projects.

The way you request V100 resources for your PBS jobs will change. Jobs will be able to request and access a subset of GPUs on the node, so each job will utilize only the GPUs it needs, leaving the rest of the GPUs on the node available to other jobs. For example, if your job needs just one GPU, you can request a single GPU when you submit the job, and the other GPUs on the node will remain available.

We highly recommend that you begin learning about this change before it is implemented, as described in the HECC Knowledge Base (KB) article Changes to PBS Job Requests for V100 GPU Resources.

Related KB articles will be updated on July 16, when the changes go into effect.

If you have any questions or concerns, please contact the Control Room at (800) 331-8737, (650) 604-4444, or support@nas.nasa.gov.