What is the H100 Cluster?
The H100 cluster is a cluster of three servers with 8 48GB H100 GPUs owned by the Data Science Institute and operated by the Center for High Throughput Computing.
Specifications:
The H100 GPU Cluster has the following specifications:
- 4x Nvidia H100 80GB HBM3 GPU
- Intel(R) Xeon(R) Platinum 8460Y+ with hyperthreading = 80 Cores
- DDR5 type Memory = 1 TB
- 13.5 TB of Enterprise NVMe Drive
Who can use the H100 Cluster?
The H100 Cluster is available to DSI Affiliates and their group members.
How do I access the H100 Cluster?
The H100 Cluster is accessed through the Center for High Throughput Computing’s HTCondor job submission process. You will first need a CHTC account. Then, by contacting us, you will receive special priority status so when you run your HTCondor jobs, they will be given priority to run on the H100 Cluster. If you qualify for an H100 account and would like priority status on the H100 Cluster, please contact Jason Lo: jason.lo@wisc.edu.
HTCondor Resources
The H100 GPU Cluster is managed using the HTCondor job scheduling system. For more information about how to get started using this system please see the links below.