Hector Martinez-Seara
HPCg Head · Scientific software
Aurum cluster
The Aurum cluster provides computing resources to all IOCB employees. Computing time is free, with no grant application needed.
Aurum is located in the basement of building B. It consists of compute nodes from four hardware generations, plus several control and storage servers. We are happy to introduce new users to high-performance computing and try to accommodate special requests, so don’t hesitate to contact us.
Ready-to-use software includes molecular dynamics (Amber, GROMACS, LAMMPS), quantum chemistry (CP2K, Gaussian, ORCA, Psi4, Turbomole, VASP), structure prediction (AlphaFold 3, ColabFold), cryo-EM (CryoSPARC, RELION, Phenix) and Schrödinger. See the full software catalog in the Aurum wiki, accessible after registration.
- ~390
- compute nodes
- ~30,000
- CPU cores
- 200+
- GPUs
- 2 PB
- storage
Hardware
The compute nodes belong to four hardware generations (a–d) that together offer
standard, high-memory and GPU nodes. You do not need to pick a node type yourself: the
scheduler routes each job to suitable nodes based on the requested time, memory and GPUs.
| Generation | Node type | Nodes | Cores / node | Memory / node | GPUs / node | Local scratch | Interconnect |
|---|---|---|---|---|---|---|---|
| a | Standard | 203 | 36 | 96 GB | 1 TB | 54 Gb/s1 | |
| a | High memory | 22 | 36 | 192 GB | 2 TB | 54 Gb/s1 | |
| a | Large memory | 7 | 36 | 768 GB | 4 TB | 54 Gb/s1 | |
| a | GPU | 5 | 36 | 96 GB | 2× Quadro RTX 5000 | 1 TB | 54 Gb/s1 |
| Total a | 237 | 8,532 cores | 10 GPUs | ||||
| b | GPU | 3 | 32 | 128 GB | 2× RTX 3080 | 2 TB | 25 Gb/s2 |
| b | GPU | 26 | 32 | 128 GB | 2× RTX 3090 | 2 TB | 25 Gb/s2 |
| b | GPU | 3 | 32 | 512 GB | 4× RTX 3090 | 2 TB | 25 Gb/s2 |
| b | Large memory | 16 | 64 | 256 GB–1 TB | 2 TB | 25 Gb/s2 | |
| b | Huge memory | 1 | 64 | 2 TB | 2 TB | 25 Gb/s2 | |
| Total b | 49 | 2,112 cores | 70 GPUs | ||||
| c | GPU | 5 | 24 | 32 GB | 2× GTX 1080 | 1 TB | 1 Gb/s2 |
| Total c | 5 | 120 cores | 10 GPUs | ||||
| d | Standard | 48 | 192 | 768 GB | 2 TB | 200 Gb/s1 | |
| d | Huge memory | 2 | 192 | 2 TB | 4 TB | 200 Gb/s1 | |
| d | GPU | 42 | 192 | 384 GB | 2× L40S | 2 TB | 200 Gb/s1 |
| d | GPU | 8 | 192 | 768 GB | 4× L40S | 2 TB | 200 Gb/s1 |
| Total d | 100 | 19,200 cores | 116 GPUs | ||||
| Aurum total | 391 | 29,964 cores | 206 GPUs |
1 Low-latency network suitable for multi-node (MPI) jobs.
2 Standard network; not suitable for multi-node jobs.