in response to feedback from the Cabeus user community to balance the demands from both development and production workloads on these limited resources, we are instituting new queue structures.

The new GPU queues for the AMD Milan A100 GPU nodes are now available for use. The queue names denote both time in hours (h) and physical node limitations (n) for each:

• gpu_12h_64n queue: Maximum submission of 12 hours and/or 64 nodes.
• gpu_24h_32n queue: Maximum submission of 24 hours and/or 32 nodes.
• gpu_36h_16n queue: Maximum submission of 36 hours and/or 16 nodes.
• gpu_devel queue: Maximum submission of 2 hours and/or 16 nodes.

Please note that beginning March 19, 2024, the gpu_normal, gpu_debug, and gpu_long queues will be disabled. Jobs submitted prior to this change will be allowed to complete.

There will be no change to the V100 GPU queues.

For more information about using the new queues, see the HECC Knowledge Base article Requesting GPU Resources.

We encourage your feedback on providing performance results and any questions or issues you have when using these queues. Please contact the Control Room at (800) 331-8737, (650) 604-4444, or support@nas.nasa.gov.