Slurm on Carina
Main content start
Slurm Queues/Partitions
Carina has three Slurm queues. The differences are whether or not GPUs can be requested, and the maximum runtime a job can have.
Most of the time, you will use --partition=normal or --partition=gpu. The long-running queue is much less common.
Standard Queues: normal and gpu
| Queue Name | CPUs | Memory | Nodes | GPUs | Max Time |
|---|---|---|---|---|---|
| normal | Dual Intel Xeon Gold 6130 (16C 2.1GHz) | 370GB | 2 | 2 Days | |
| gpu | Dual Intel Xeon Gold 6130 (16C 2.1GHz) | 350GB | 2 | Dual P100 | 2 Days |
| gpu | Dual Intel Xeon Silver 4114 (10C 2.2GHz) | 250GB | 5 | Quad NVIDIA Tesla V100 | 2 Days |
| gpu | Dual Intel Xeon Gold 6330 (56C 2GHz) | 250GB | 4 | Quad NVIDIA A100 with 14TB of local scratch | 2 Days |
Specialty Queue: gpu-long
| Queue Name | CPUs | Memory | Nodes | GPUs | Max Time |
|---|---|---|---|---|---|
| gpu-long | Dual Intel Xeon Silver 4114 (20C 2.2GHz) | 250GB | 2 | Quad NVIDIA Tesla V100 | 5 Days |
| gpu-long | Dual Intel Xeon Gold 6330 (56C 2GHz) | 250GB | 2 | Quad NVIDIA A100 with 14TB of local scratch | 5 Days |