POD

64 node (~2,600 core) dual Intel 6148 CPUs with OmniPath interconnect.

Node[1-55,57-63] dual 20 core 6148s - regular compute node / batch queue
Node 56 and 64 - regular compute node / short queue (2 hour time limit for jobs) [node55 is not always up]
Node[101-104] - large memory compute nodes (1 TB RAM) dual 20 core 6148s / largemem queue
Node[111-113,116-125] - GPU nodes - 4 NVIDIA 64-bit V100 GPUs connected with NVLink per node / gpu queue

pod-gpu - GPU development node
/scratch - 19TB SSD RAID0 scratch (not backed up)
/bigscratch - 70TB NVMe RAID0 scratch (not backed up)