GPU Servers

• Systems offers a small handful of servers equipped for robust GPU analysis for things like visualization, large-scale tensor operations, deep learning, or parallelizable workloads.

• Currently, the GPU servers available include:

  • pixie.dfci.harvard.edu (155.52.47.158) : 4 NVIDIA GTX Titan X

  • aries.dfci.harvard.edu (155.52.47.210) : 4 NVIDIA A100-80GB

  • cetus.dfci.harvard.edu (172.24.216.81) : 2 NVIDIA A100-80GB

  • polaris.dfci.harvard.edu (155.52.47.161):2 NVIDIA H100-95GB

  • fornax.dfci.harvard.edu (172.24.216.91) : 2 NVIDIA A800-40Gb ,2 NVIDIA A100-40Gb

• You can connect to any of the servers above either from a login server such as ada / noah, or directly from your own machine, as long as you're connected to the internal network via VPN / on-site wifi.

• If you have a question when using these systems, it's possible someone else has had the same one before. Please feel free to ask a peer or systems admin in the #gpus Slack channel.

Note: For guidance on connecting to any of our Unix-based servers, please refer to our instructions on this page.

Please follow the community-written guidelines below, co-authored entirely by the faculty & staff that you will be sharing the resources with!