NVIDIA Gracehopper Node Checkout

NVIDIA Gracehopper Boxes

UW ARCC has two NVIDIA Gracehopper boxes allowing users the ability to work with the NVIDIA ARM processors and GH200 GPUs on our HPC cluster. These are nodes are standalone servers with ARM based CPU architecture available for reservation by UW researchers. These nodes run an NVIDIA provided OS image containing the NVIDIA GPU toolset. This OS is based on the Ubuntu 22.04.2 release.

These machines are experimental nodes that share system and GPU memory.

Per the NVIDIA Site:

“The NVIDIA Grace Hopper Superchip architecture brings together the groundbreaking performance of the NVIDIA Hopper GPU with the versatility of the NVIDIA Grace CPU, connected with a high bandwidth and memory coherent NVIDIA NVLink Chip-2-Chip (C2C) interconnect in a single superchip, and support for the new NVIDIA NVLink Switch System… Traditional heterogeneous platforms with PCIe-connected accelerators require users to follow a complex programming model that involves manually managing device memory allocations and data transfer to and from the host.”

The NVIDIA Grace Hopper Superchip platform is heterogeneous and easy to program, and NVIDIA is committed to making it accessible to all developers and applications, independent of the programming language of choice.”

To request access, fill out a new project request form on our portal selecting “Beartooth” as the service and checking the box for “NVIDIA Gracehopper Node Reservation” under the specialized services options.
If you have a timeframe for your reservation, please include those values in your project request under project start date and project completion date.

Please plan to make reservation requests 1-2 weeks before the start date of your reservation time.

Node reservations must be confirmed and approved by ARCC staff. Approved reservations will be honored based on the start/end time requested within the project request form.