System Technical Specification

Hardware Details for the GM4 Cluster

Cluster

  • 29 GPU compute nodes each with 4 Nvidia V100 GPUs
  • 4 CPU only compute nodes

Compute Nodes

Each node has the following base components:

  • 2 Intel Skylake 6148  processors (40 cores per node)
  • 192 GB of conventional memory
  • 480 GB of local SSD storage
  • EDR Infiniband (100 Gbps) network card

Storage

Distributed file system storage: 500 TB of high performance GPFS storage  

Local storage: 480 GB of local storage on each node  

Network

EDR (100 Gbps)  Infiniband fabric between nodes and with the storage appliance in a fat tree network topology 

GPU Card Specs

            System Diagram