SMP Cluster

The SMP cluster is designed for workloads that run on a single server using shared memory parallelism. Each node provides multiple CPU cores with access to a common memory space, making the cluster well suited for multithreaded applications, OpenMP codes, and jobs that do not require distributed computing across multiple nodes.

The cluster has two partitions. Most jobs run on the default smp partition. Jobs that need more memory than those nodes provide can use the high-mem partition, whose nodes offer up to 2 TB of RAM on a single node. The two partitions are billed at different rates (the high-mem nodes cost more per core but less per GB of memory) — see Service Units for the exact weights.

Specifications

Nodes are grouped by partition, newest hardware first.

Partition Nodes --constraint CPU Cores/Node Mem/Node Mem/Core Scratch Network Node Names
smp 1 amd,turin AMD EPYC 9755 256 1.5 TB 6 GB 3.2 TB NVMe 10GbE smp-n266
smp 38 amd,genoa AMD EPYC 9374F 64 768 GB 12 GB 3.2 TB NVMe 10GbE smp-n[214-251]
smp 55 amd,rome AMD EPYC 7302 32 256 GB 8 GB 960 GB NVMe 10GbE smp-n[156-210]
high-mem 2 intel,ice_lake Intel Xeon Platinum 8352Y 64 2 TB 32 GB 10.24 TB NVMe 10GbE smp-2048-n[0-1]
high-mem 8 intel,ice_lake Intel Xeon Platinum 8352Y 64 1 TB 16 GB 10.24 TB NVMe 10GbE smp-1024-n[1-8]
high-mem 1 amd,naples AMD EPYC 7351 32 1 TB 32 GB 1 TB NVMe 10GbE smp-1024-n0

Additional Features

To request a particular feature (such as an Intel host CPU), add the following directive to your job script:

#SBATCH --constraint=intel

Multiple features can be requested by providing a comma-separated list (without intervening spaces):

#SBATCH --constraint=amd,genoa

  • What a job costs

    How Service Units are calculated — the smp and high-mem partitions bill differently.

    Service Units

  • Request resources

    Choose cores, memory, and a partition for an interactive or batch job.

    Requesting Resources

  • Limits & priority

    Per-group CPU and memory limits, QoS walltimes, and how priority is computed.

    Job Scheduling Policy