ClinWulf is a HIPAA-compliant high-performance computing (HPC) cluster designed to enable Stony Brook Medicine researchers to analyze clinical data utilizing NVIDIA GPUs. Requests for access requires Principal Investigator or Supervisor approval, CITI and HIPAA training completion, an executed Data Use Access agreement, and most importantly, UHMC credentials.
- Consists of 27 backend nodes – 20 CPU only nodes and 7 GPU nodes
- Intel Xeon Gold 6148 2.4 GHz (Skylake family), CPUs with dual sockets providing the total 1,040 CPU cores with access to 8,548 GB total DDR4 memory
- 24x NVIDIA GeForce RTX 2080 Ti each with 11 GB onboard GPU RAM providing the total 104,448 cores and theoretical performance of 322.8 TFLOPS of single-precision floating points in 6 GPU nodes
- 18 TB of local NVMe SSDs on 6 GPU nodes
- 8 NVIDIA A100 GPU cards linked together via NVLink aggregating all eight GPU cards and providing a block of 320 GB of GPU RAM along with 55,296 total cores and 3,456 tensor cores in one DGX GPU node
- 4 AMD EPYC 7742 64-Core processor giving additional 256 CPU cores at 2.25 GHz per core in DGX GPU node
- 1 TB DDR4 memory and 14 TB of NVMe SSD available in DGX GPU node
- 1 PB of central storage for the HPC cluster
- Mellanox InfiniBand switch interconnecting all nodes