Head over to the RC Dashboards to see our SPORC cluster in action!

You can also take a look at our cluster benchmarking results.

SPORC Cluster: (Scheduled Processing On Research Computing)

  • 2304 cores (Intel(R) Xeon(R) Gold 6150 CPU @ 2.70GHz)
  • Linpack benchmark, cpu-only: 105 TFLOPS
  • 24TB RAM
  • 100 Gbit/sec RoCEv2 interconnect (Mellanox MLX5/Juniper QFX5210-64c)
  • 64 SuperMicro X11 systems with space for 4 GPU cards
  • 100 Nvidia A100 cards
    • 40GB HBM2
    • 156 TFLOPS TF32
    • 312 TFLOPS FP16
    • 624 TOPS Int8
  • 16 Nvidia V100 cards
    • 32GB HBM2
    • 14 TFLOPS single-precision
    • 7 TFLOPS double-precision
    • 56 TOPS Int8
    • 112 TensorFLOPS
  • 96 Nvidia P4 cards
    • 8GB GDDR5
    • 5.5 TFLOPS single-precision
    • 22 TOPS Int8
  • Interactive Partition:
    • 148 cores
    • 525GB RAM
    • 2 Nvidia A100 cards
    • 42 Nvidia P4 cards

Ceph (Block and Filesystem) Storage

  • 9.5PB of storage comprised of
  • 200G NVMe hot tier
  • 6PB 7200RPM SAS bulk tier
  • 3.5PB 7200RPM SATA cold tier
  • 15TB Intel(R) Optane(R) NVMe CephFS metadata tier
  • Dual 100 Gbit/sec RoCEv2 interconnect (Mellanox MLX5/Juniper QFX5210-64c)

Large Single-System-Image Compute Node (theocho)

  • SYS-7089P Supermicro 8-way
  • 144 cores (Intel(R) Xeon(R) Platinum 8168 CPU @ 2.70GHz)
  • Linpack benchmark, cpu-only: 4 TFLOPS
  • 2.3TB RAM
  • 1 Nvidia V100 card

Operating System

  • Red Hat Enterprise 7