The following page contains hardware specifications for each of the Scalable Units, in the Discover cluster.
Scalable Unit 16 – CPU-only Nodes
- Manufacturer: Aspen Systems Inc
- 1,912 Tflop/s
- 32,448 Intel Xeon Cascade Lake Refresh processor cores
- 676 Supermicro TwinPro nodes
- 2 24-core processors per node
- 192 GB of memory per node
- No Swap Space
- 2.4 GHz Intel Xeon Cascade Lake Refresh
- Interconnect: HDR100 Infiniband 100 Gbps
- Production: 3Q 2021
Scalable Unit 16 – CPU & GPU Nodes
- Manufacturer: Aspen Systems Inc
- 12 Supermicro GPU nodes, each with AMD EPYC Rome and 4 NVIDIA A100 GPUs
- Four NVIDIA A100 GPUs per node, each with
- 9.7 Dual precision TFLOP/s
- 6,912 CUDA cores
- 40 GB memory
- 600 GB/s NVLINK GPU to GPU connection
- AMD EPYC Rome 48-core nodes:
- 516 Tflop/s total
- 576 total AMD EPYC Rome processor cores
- 2 24-core processors per node
- 512 GB of memory per node
- 2.8 GHz AMD EPYC Rome
- No Swap Space
- Interconnect: Dual HDR Infiniband 2×200 Gbps
- Four NVIDIA A100 GPUs per node, each with
- Production: 3Q 2021
Scalable Unit 17 – CPU-only Nodes
- Manufacturer: Aspen Systems Inc
- 2,883 Tflop/s
- 90,112 AMD Milan EPYC processor cores
- 704 Supermicro TwinPro nodes
- 2 64-core processors per node
- 512 GB of memory per node
- No Swap Space
- 2.0 GHz AMD Milan EPYC
- Interconnect: HDR100 Infiniband 100 Gbps
- Production: 4Q 2023
Scalable Unit 18 – CPU-only Nodes
- Manufacturer: Aspen Systems Inc
- 2,883 Tflop/s
- 90,112 AMD Milan EPYC processor cores
- 704 Supermicro TwinPro nodes
- 2 64-core processors per node
- 512 GB of memory per node
- No Swap Space
- 2.0 GHz AMD Milan EPYC
- Interconnect: HDR100 Infiniband 100 Gbps
- Production: 2Q 2024