Resources

Cluster Overview

Metric

Value

Compute Nodes

580+

GPU Nodes

74

Total CPU Cores

30,000+

Total Memory

350 TB+

GPUs

200 P100, 4 V100, 16 H100, 64 H200, 16 L40S

Storage

4 PB VAST

Operating System

Red Hat Enterprise Linux 9

Interconnect

EDR InfiniBand (most nodes); NDR InfiniBand on L40S and H200 GPU nodes

Login Nodes

Standard Login

Host: login.hpc.caltech.edu (Caltech network only)

Specification

Value

Nodes

2

Cores per node

64 (2× 32-core Intel Icelake @ 2.2 GHz)

Memory per node

527 GB

Visualization Login

Host: vislogin.hpc.caltech.edu (Caltech network only, web-based via Open OnDemand)

Specification

Value

Nodes

2

Cores per node

32 (2× 16-core Intel Cascadelake @ 2.1 GHz)

Memory per node

192 GB

Compute Nodes

All compute nodes have a 7-day walltime limit.

Type

Nodes

Total Cores

Total Memory

Cores / Node

CPU

Memory / Node

Standard Compute

179

5,728

34 TB

32

2× 16-core Intel Skylake @ 2.1 GHz

192 GB

Large Memory

11

352

4 TB

32

2× 16-core Intel Skylake @ 2.1 GHz

384 GB

1st Expansion

85

4,760

32 TB

56

2× 28-core Intel Cascadelake 8276 @ 2.1 GHz

386 GB

Expansion Big Memory

9

504

13.5 TB

56

2× 28-core Intel Cascadelake 8276 @ 2.2 GHz

1.5 TB

Expansion Big Memory 4-Way

2

192

3 TB

96

4× 24-core Intel Cascadelake 8276 @ 2.4 GHz

1.5 TB

2nd Expansion

61

3,904

32 TB

64

2× 32-core Intel Icelake 8352Y @ 2.2 GHz

512 GB

GPU Nodes

All GPU nodes have a 7-day walltime limit.

Type

Nodes

Total Cores

Total Memory

Cores / Node

CPU

GPUs (per node)

Memory / Node

NVIDIA P100

50

1,400

12.8 TB

28

2× 14-core Intel Broadwell @ 2.4 GHz

4× P100 / 16 GB

256 GB

NVIDIA V100

2

56

500 GB

28

2× 14-core Intel Broadwell @ 2.4 GHz

4× V100 / 16+32 GB

256 GB

NVIDIA H100

4

512

6 TB

128

2× 64-core AMD EPYC 9554 @ 3.1 GHz

4× H100 / 80 GB

1.5 TB

NVIDIA L40S

4

512

3 TB

128

2× 64-core AMD EPYC 9554 @ 3.1 GHz

4× L40S / 46 GB

768 GB

NVIDIA H200

18

2,304

27 TB

128

2× 64-core AMD EPYC 9554 @ 3.1 GHz

4× H200 / 144 GB

1.5 TB

See also

GPU Computing for partition selection, typed --gres=gpu:<type>:N syntax, and CUDA/driver guidance.

Contact

For detailed specifications or custom requirements, contact help-hpc@caltech.edu.