MGHPCC GPU Cluster
Intro to MGHPCC Cluster
The MGHPCC cluster provides a low cost computing footprint for starting on AI projects.
The cluster is comprised of a rack of Lenovo SR675 servers interconnected with NVIDIA QM9790 64 Port Quantum NDR InfiniBand Switch for fast internode communication and access to high-performance storage.
A high-performance Weka filesystem is deployed in converged mode across the cluster for a total of 75.43 TiB total usable (expandable to 150TiB). This storage is snapped to NESE object store on the backend with a lifecycle policy to migrate data between tiers.
Node Specs
- Lenovo ThinkSystem SR675 V3
- 2x AMD Epic 9334 processors: 64 cores
- 786GB DDR5 Memory
- 8x NVidia L40S 48GB PCIe Gen4 GPUs
- 4x 15.36TB Read Intensive NVMe