SeaWulf GPU Nodes

SeaWulf GPU Nodes Guide

GPU nodes on SeaWulf provide access to Haswell, Skylake, and AMD Milan nodes with A100, P100, or V100 GPUs. These nodes are designed for GPU-accelerated computing workloads. A small subset of node memory is reserved for the OS and filesystem.

Available GPU Queues

Queue CPU Architecture Vector/Matrix Extension CPU Cores per Node GPUs per Node Node Memory Default Runtime Max Runtime Max Nodes Min Nodes Max Simultaneous Jobs per User Multiple Users per Node
gpu Intel Haswell AVX2 28 4 128 GB 1 hour 8 hours 2 n/a 2 No
gpu-long Intel Haswell AVX2 28 4 128 GB 8 hours 48 hours 1 n/a 2 No
gpu-large Intel Haswell AVX2 28 4 128 GB 1 hour 8 hours 4 n/a 1 No
p100 Intel Haswell AVX2 12 2 64 GB 1 hour 24 hours 1 n/a 1 No
v100 Intel Haswell AVX2 28 2 128 GB 1 hour 24 hours 1 n/a 1 No
a100 Intel Ice Lake AVX512 & Intel DL Boost 64 4 256 GB 1 hour 8 hours 2 n/a 2 Yes
a100-long Intel Ice Lake AVX512 & Intel DL Boost 64 4 256 GB 8 hours 48 hours 1 n/a 2 Yes
a100-large Intel Ice Lake AVX512 & Intel DL Boost 64 4 256 GB 1 hour 8 hours 4 n/a 1 Yes

Instructions for Using GPU Nodes

  • Select the appropriate queue: Choose based on GPU type, number of GPUs, node memory, and expected runtime.
  • Request memory explicitly: Use #SBATCH --mem=[amount] to ensure your job stays within limits, especially on shared nodes.
  • Request GPUs: Use #SBATCH --gpus=[number] to specify GPU allocation.
  • Monitor GPU usage: Use nvidia-smi inside your job or interactive session.
  • Respect other users: Shared nodes require careful memory and GPU management to prevent interference.
  • Adhere to runtime limits: Do not exceed the maximum runtime listed for each queue.

Note on NVwulf Access

If you are having trouble accessing GPUs on these queues or need additional dedicated GPU resources, you may request access to the NVwulf cluster. NVwulf provides extra GPU resources for high-demand workloads. Submit a request through the HPC portal to gain access.