Have a Question?
Hardware
HPC/AI Cluster Specifications
ARCC2 – Discovery Cluster
- 94x HPE CPU compute nodes:
- 2x AMD EPYC 7452 CPU (32 Cores, 3.35 GHz)
- 256 GB RAM (DDR4-3200)
- 1x HPE Large memory CPU compute node:
- 2x AMD EPYC 7452 CPU (32 Cores, 3.35 GHz)
- 2048 GB RAM (DDR4-3200)
- 2x HPE V100 GPU compute node:
- 2x AMD EPYC 7452 CPU (32 Cores, 3.35 GHz)
- 256 GB RAM (DDR4-3200)
- 2x NVIDIA V100 PCIE 32GB GPU
- 10x HPE A100 GPU compute nodes:
- 2x AMD EPYC 7452 CPU (32 Cores, 3.35 GHz)
- 1024 GB RAM (DDR4-3200)
- 2x NVIDIA A100 PCIE 40GB GPU
- 1x HPE HGX H100 GPU compute node:
- 2x Intel Xeon Platinum 8468 CPU (48 Cores, 3.8 GHz)
- 1024 GB RAM (DDR5-4800)
- 8x NVIDIA H100 SXM 80GB GPU
- 2x Infiniband HDR 200 Gb/s ea.
- 1x HPE RTX PRO 6000 SE GPU compute node:
- 2x Intel Xeon 6767P CPU (64 Cores, 3.9 Ghz)
- 512 GB RAM (DDR5-6400)
- 4x NVIDA RTX PRO 6000 Server Edition PCIE 96GB GPU
- 2x Infiniband NDR 400 Gb/s ea.
- 1x HPE GPU H200 compute node:
- 2x AMD EPYC 9355 (32 Cores, 4.4 GHz)
- 1512 GB RAM (DDR5-6400)
- 8x NVIDIA H200 SXM 141GB GPU
- 2x Infiniband NDR 400 Gb/s ea.
- Mellanox InfiniBand HDR 200 Gb/s networking fabric
ARCC1 – Pilot cluster donated to Open Science Grid Consortium – OSPool
- 36x Dell CPU compute nodes each configured with:
- 2x Intel Xeon Gold 6148 CPU (20 Cores, 2.4-3.7 GHz)
- 192 GB RAM
- Intel OmniPath 100Gb/s networking fabric
ARC Storage Services
| Storage Type | Use Case | Backed up (Y/N) |
|---|---|---|
| Scratch | High-speed Lustre storage intended for data sets in active use by the HPC cluster | N |
| Project | High-speed Lustre storage intended for data sets in active use by the HPC cluster | Y |
| Bulk | Intended for large data volume that needs to be analyzed/processed in the near future | Y |
Note: Scratch storage is limited to 20TB and is purged on a regular basis.
For more information on storage, please visit the ARC Storage Services knowledgebase article. For questions, please email us.