St. Jude HPC Research Cluster

St. Jude's own research-computing page describes its on-campus HPC environment as four centralized systems: an HPC Research Cluster of more than 300 general nodes (25,000 cores) plus about 60 GPU nodes with over 400 GPUs installed; a PCAI AI/ML inference cluster; a Slurm AI/ML training cluster built from 14 HPE Cray XD670 nodes (112 CPU cores, 8 NVIDIA H100 SXM 80GB GPUs and 1024GB RAM per node, InfiniBand with GPUDirect RDMA/GDS); and a VR/Vis cluster. The page states the overall HPC environment provides '5.3 PFLOPs of compute power and over 150 PB research storage capacity.' This row records that aggregate figure as rpeak_pflops (operator_stated, since it is a hospital-published figure rather than an HPL benchmark result) and the 25,000-core count from the general HPC Research Cluster nodes only, not the separate GPU/training clusters, because St. Jude's own page never totals all clusters into one core count. No second, independently reporting source with matching hardware detail was found (an HPCwire feature on St. Jude's HPC/AI programme covers the same broad story but restates no specific node, core or GPU figures), so confidence stays reported. The page's description of the PCAI cluster is internally inconsistent between '10 DGX A100 worker nodes' and 'HPE ProLiant DL380a' server specs in different places, so PCAI hardware is deliberately left out of system_components to avoid asserting an uncertain configuration.