A CPU cluster composed of 88 nodes, with a total of 12,584 CPU cores and 99.8TB of memory. This cluster supports large-scale computation, low-latency networking for MPI workloads, large memory systems, and high-performance parallel filesystems. 

Hardware

CPU Compute Nodes 

•  The node500 class comprises 16 nodes with two AMD EPYC 9655 CPUs (192 cores per node), 1.5 TB of RAM, 100 Gb Ethernet, and 200 Gb InfiniBand connectivity. 

• The node400 class comprises 21 nodes with two AMD EPYC 9654 CPUs (192 cores per node), 1.5 TB of RAM, 25 Gb Ethernet, and InfiniBand connectivity. 

• The node300s class comprises 39 nodes, each with two AMD EPYC 7763 CPUs (128 cores per node), 1 TB of RAM, 25 Gb Ethernet, and InfiniBand connectivity. 

• The node200 class comprises 9 nodes, each with two Intel Xeon Gold 6230 CPUs (40 cores per node), 96 GB of RAM, 10 Gb Ethernet, and no InfiniBand. 

• The Large memory node is one node with two AMD EPYC 7543 CPUs (64 cores total), 4 TB of RAM, 25 Gb Ethernet. 

• The high clock class consists of 2 nodes with two AMD EPYC 7F52 CPUs (32 cores per node), 256 GB of RAM, 25 Gb Ethernet, and InfiniBand connectivity. 

Software

• Operating System: RHEL 9.6, with GNU compilers (gcc, Fortran)
• Resource Manager: Slurm version 25.11.2
• Package Manager: Lmod
• CUDA 13

A GPU cluster comprised of 104 Nvidia H200 GPUs and 48 Nvidia Pro 6000 GPUs. Nodes have 100 Gb Ethernet and 200 Gb InfiniBand, as well as 1.0 TB of RAM and 128 or 192 cores, depending on the server (H200 nodes vs. Pro 6000 nodes, respectively). 

Hardware

GPU Compute Nodes

• The IceCore class comprises 16 nodes, each with four NVIDIA H200 NVL 141 GB GPUs, two Intel Xeon 6740E CPUs (192 cores per node), 1.0 TB of RAM, 100 Gb Ethernet, and 200 Gb InfiniBand connectivity. It also comprises 5 nodes, each with eight Nvidia H200 SXM 141 Gb GPUs, two Intel Xeon Platinum 8593Q CPUs (128 cores per nodes), 1.0 TB of RAM, 100 Gb Ethernet, and 200 Gb InfiniBand connectivity and also 6 nodes, each with eight Nvidia RTX Pro 6000 96 Gb GPUs, two Intel Xeon 6740E CPUs (192 cores per node), 1.0 TB of RAM, 100 Gb Ethernet, and 200 Gb InfiniBand connectivity. 

• The H100 class comprises 2 nodes, each with four NVIDIA H100 80 GB GPUs, two Intel Xeon Platinum 8462Y+ CPUs (64 cores per node), 1 TB of RAM, 100 Gb Ethernet, and InfiniBand connectivity. 

• The A100 class comprises 2 nodes, each with two NVIDIA A100 40 GB GPUs, two AMD EPYC 7763 CPUs (128 cores per node), 1 TB of RAM, 25 Gb Ethernet, and InfiniBand connectivity. 

Software

• Operating System: RHEL 9.6, with GNU compilers (gcc, Fortran)
• Resource Manager: Slurm version 25.11.2
• Package Manager: Lmod
• CUDA 13

A large memory, sharded MongoDB to support near real-time access to enormous data files. This supports projects that require such speed to analyze, describe, and explain rapidly growing datasets effectively.  Comprised of 8 nodes, each with two Intel Xeon 6348 CPUs (56 cores per node), 8 TB of RAM, and 25 Gb Ethernet.

Will Be Available Summer 2026

Three GPU Nodes, each equipped with an NVIDIA L40S GPU, an AMD EPYC 9354P, and 192GB of RAM. One Large Memory Node with an AMD EPYC 9554P and 1.5TB of RAM. Two Proxmox Nodes, each equipped with an AMD EPYC 9354P and 384GB of RAM.

IBM Spectrum Scale (GPFS) filesystems are available for storing research data and software during computation. These filesystems provide 2.8 PB of storage, with roughly 25% of capacity on NVMe/Flash storage devices for performance.

The VACC provides 100Gb, 40Gb, 25Gb, and 10Gb Ethernet connections to different portions of the cluster.

4 I/O nodes with 100GbE, 200Gb HDR InfiniBand providing access to: 

• IBM FS7300 provides 664TB of Flash Storage, and an expansion drawer provides 754TB of spinning disk storage. 

• Dell MD3460 with expansion drawer, providing 564TB of disk storage. 

• Dell ME4084 providing 682 TB of spinning-disk storage. 

• IBM FS7200 providing 187TB of NVMe-attached FlashCore Module storage. 

• IBM FS5600 providing  633TB of NVMe-attached FlashCore Module storage plus 1.1PB of spinning disk storage.