The Andromeda HPC Cluster offers enterprise grade compute and data storage for your research needs. Initially installed in the Fall of 2020, Andromeda recently (2025) underwent a configuration overhaul (January) and resource expansion (June).
Environment
- RedHat EL 9
- Slurm 23.11
- OFED 24.10
CPU Nodes
Node Count | Access | CPU Type | CPUs/node | Memory (RAM) |
92 | Community | Intel Xeon Platinum 8260 @2.4Ghz | 48 | 192GB |
134 | Community | Intel Xeon Platinum 8352Y @2.2Ghz | 64 | 256GB |
20 | Community | Intel Xeon Platinum 8568Y+ @2.3Ghz | 96 | 2TB |
6 | Private | Intel Xeon Platinum 8452Y @2.0Ghz | 72 | 512GB |
6 | Private | Intel Xeon Platinum 8568Y+ @2.3Ghz | 96 | 512GB |
GPU Nodes
Node Count | Access | GPU Model | Memory(VRAM) | GPUs/node | CPU Type | CPUs/node | Memory (RAM) |
3 | Community | Nvidia A100 | 40GB | 4 | Intel Xeon Platinum 8362 @2.8Ghz | 64 | 256GB |
4 | Community | Nvidia A100 | 80GB | 4 | Intel Xeon Platinum 8358 @2.6Ghz | 64 | 256GB |
2 | Community | Nvidia V100 | 64GB | 4 | Intel Xeon Platinum 8260 @2.4Ghz | 48 | 192GB |
1 | Private | Nvidia V100 | 64GB | 4 | Intel Xeon Platinum 8260 @2.4Ghz | 48 | 192GB |
2 | Private | Nvidia A10 | 96GB | 4 | Intel Xeon Platinum 8362 @2.8Ghz | 64 | 512GB |
1 | Private | Nvidia L40S | 48GB | 4 | Intel Xeon Platinum 8452Y @2.0Ghz | 64 | 512GB |
1 | Private | Nvidia L4 | 24GB | 7 | Intel Xeon Platinum 8562Y @2.8Ghz | 64 | 256GB |
10 | Community | Nvidia L40S | 48GB | 4 | Intel Xeon Platinum 8562Y @2.8Ghz | 64 | 512GB |
1 | Private | Nvidia H100 | 94GB | 2 | Intel Xeon Platinum 8562Y @2.8Ghz | 64 | 512GB |
1 | Community | Nvidia H200 | 141GB | 8 | Intel Xeon Platinum 8562Y @2.8Ghz | 64 | 2TB |
Node Totals
Nodes | CPU(cores) | CPUs (RAM) | GPUs | GPUs (VRAM) |
258 (cpu) | 15,920 | 99,072GB (~99TB) | NP | NP |
26(gpu) | 1,616 | 11,840GB (~11.8TB) | 109 | 6,892GB (~6.9TB) |
284 (total) | 17,536 | 110,912GB (~111TB) | 109 | 6,892GB (~6.9TB) |
Network
1Gbps Management/SSH between compute nodes
10Gbps Ethernet to Campus and the Internet
200Gbps HDR Infiniband between compute nodes