Partitions
Slurm partitions are separate queues that divide up a cluster's nodes based on specific attributes. Each partition has its own constraints, which control which jobs can run in it.
| Partition | Description | Per-node Limits | Max Walltime |
|---|---|---|---|
| common | Jobs that will run CPU only. | CPUs: 128 RAM: 500G |
7 days |
| common-lm | Large memory CPU-only jobs on the idle CPUs and memory on the GPU nodes. Jobs may be pre-empted (cancel and requeue) by jobs in any GPU partition. | CPUs: 96 RAM: 960G |
7 days |
| interactive | Short term CPU jobs with a max walltime of 1 hour. Higher priority than common. |
CPUs: 128 RAM: 500G |
1 hour |
| gpu | Jobs that will run on H200 GPU nodes through general access. Jobs may be pre-empted (cancel and requeue) for higher priority jobs. Contact your institutional representative to gain access. | CPUs: 192 RAM: 1920G |
2 days |
| interactive-gpu | Short term GPU jobs with a max walltime of 1 hour. Contact your institutional representative to gain access. Higher priority than gpu. |
CPUs: 192 RAM: 1920G |
1 hour |
| gpu-hp | Jobs that will run on H200 GPU nodes with higher priority. Restricted to partner institutions. Each institution is allocated a monthly pool of GPU-hours. Contact your institutional representative to gain access. | CPUs: 192 RAM: 1920G |
7 days |
| workshop | Temporary partition for workshop and training events. Access is by request only. | - | - |
Note
If a partition is not specified, the default partition is the common partition.
Note
CPUs are hardware threads (2 per physical core). On common-lm and the GPU partitions, memory is capped at 10 GB per CPU, and Slurm raises the CPU count of a job that asks for more, e.g., --mem=960G is given 96 CPUs. The most a single-node job can request is therefore 960 GB on common-lm and 1,920 GB on the GPU partitions. --hint=nomultithread halves both, because each CPU then occupies a whole core. See Partition resource totals for how to use every physical core without that penalty.
Tip
To submit to more than one partition, give -p or --partition a comma-separated list. Slurm starts the job in whichever partition can run it first. E.g., --partition common,common-lm
Guidance on GPU usage can be found in the GPU documentation.