Cluster Specifications
Table of Contents:
Related Articles:
CPU and GPU Specifications
There are 145 nodes in the chip cluster divided into partitions. Please refer to this page for more information on the partitions: chip Partitions and Usage
CPU Nodes
49 nodes purchased in 2018 that each have 2x 18-core Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz (Skylake) with 358 GB of reservable CPU memory and 72GB of local storage;
18 nodes purchased in 2021 that each have 2x 24-core Intel(R) Xeon(R) Gold 6240R CPU @ 2.40GHz (Cascade Lake) with 178 GB of reservable CPU memory and 396GB of local storage;
13 nodes purchased in 2024 that each have 2x 32-core Intel(R) Xeon(R) Gold 6548Y+ (Emerald Rapids) with 931 GB of reservable CPU memory and 1.8TB of local storage;
38 nodes purchased in 2024 that each have 2x 32-core Intel(R) Xeon(R) Gold 6548Y+ (Emerald Rapids) with 465 GB of reservable CPU memory and 1.8TB of local storage.
CPU Node Schematics: Node Schematics
GPU Nodes
4 nodes purchased in 2020, with 2x 24-core Intel(R) Xeon(R) Gold 6240R CPU @ 2.40GHz (Cascade Lake) with 376 GB of reservable CPU memory, 8x RTX 2080Ti GPU cards each with 11 GB of GPU memory, and 1.9TB;
7 nodes purchased in 2020, with 2x 24-core Intel(R) Xeon(R) Gold 6240R CPU @ 2.40GHz (Cascade Lake) with 376 GB of reservable CPU memory, 8x RTX 6000 GPU cards each with 24 GB of GPU memory, and 1.9TB;
2 nodes purchased in 2020, with 2x 24-core Intel(R) Xeon(R) Gold 6240R CPU @ 2.40GHz (Cascade Lake) with 753 GB of reservable CPU memory, 8x RTX 8000 GPU cards each with 48 GB of GPU memory, and 1.9TB;
2 nodes purchased in 2024, with 2x 16-core Intel(R) Xeon(R) Gold 6548Y+ (Emerald Rapids) with 251 GB of reservable CPU memory, 2x H100 GPU cards each with 100 GB of GPU memory and NVLink between them, and 7TB of local storage;
10 nodes purchased in 2024, with 2x 16-core Intel(R) Xeon(R) Gold 6548Y+ (Emerald Rapids) with 251 GB of reservable CPU memory, 4x L40S GPU cards each with 48 GB of GPU memory, and 7TB of local storage.
GPU Node Schematics: Node Schematics
This information is summarized in the table below:
Year Purchased | CPU Cores | CPU Mem | GPU Cards | GPU Mem/Card | Scratch Space | Number | CPU Arch |
2018 | 36 | 376 GB | 0 | N/A | 69G | 49 | |
2020 | 48 | 384 GB | 8 (RTX 2080Ti) | 11 GB | 1.9T | 4 | |
2020 | 48 | 384 GB | 8 (RTX 6000) | 24 GB | 1.9T | 7 | |
2020 | 48 | 768 GB | 8 (RTX 8000) | 48 GB | 1.9T | 2 | |
2021 | 48 | 187 GB | 0 | N/A | 397G | 18 | |
2024 | 64 | 1024 GB | 0 | N/A | 1.8T | 13 | |
2024 | 64 | 512 GB | 0 | N/A | 1.8T | 38 | |
2024 | 32 | 256 GB | 2 (H100) | 100GB | 7.0T | 2 | |
2024 | 32 | 256 GB | 4 (L40S) | 48GB | 7.0T | 10 |
Operating System
All nodes are running the RHEL 9.5.
Networking Connections
The CPU cluster nodes are all connected to each other and enterprise storage by a high speed backend Infiniband network supporting 100 Gbps. The GPU cluster nodes are all connected to each other and enterprise storage by a high speed backend Ethernet network supporting 10 Gbps.