Trillium: Difference between revisions
Jump to navigation
Jump to search
mNo edit summary |
mNo edit summary |
||
| Line 22: | Line 22: | ||
* 1,224 CPU nodes, each with | * 1,224 CPU nodes, each with | ||
** Two 96-core AMD EPYC “Zen5” processors (192 cores per node). | ** Two 96-core AMD EPYC “Zen5” processors (192 cores per node). | ||
** 768 GiB of memory. | ** 768 GiB of DDR5 memory. | ||
* 60 GPU nodes, each with | * 60 GPU nodes, each with | ||
** 4 x NVIDIA H100 SXM 80GB | ** 4 x NVIDIA H100 SXM 80GB | ||
** One 96-core AMD EPYC “Zen4” | ** One 96-core AMD EPYC “Zen4” processor. | ||
** 768 GiB of memory. | ** 768 GiB of DDR5 memory. | ||
* Nvidia “NDR” Infiniband network | * Nvidia “NDR” Infiniband network | ||
Revision as of 04:19, 5 November 2024
| Availability: Spring 2025 |
| Login node: to be determined |
| Globus endpoint: to be determined |
| Data transfer node (rsync, scp, sftp,...): to be determined |
| Portal: to be determined |
This is the page for the large parallel cluster named Trillium hosted by SciNet at the University of Toronto.
The Trillium cluster will be deployed in the spring of 2025.
This cluster, built by Lenovo Canada, will consist of:
- 1,224 CPU nodes, each with
- Two 96-core AMD EPYC “Zen5” processors (192 cores per node).
- 768 GiB of DDR5 memory.
- 60 GPU nodes, each with
- 4 x NVIDIA H100 SXM 80GB
- One 96-core AMD EPYC “Zen4” processor.
- 768 GiB of DDR5 memory.
- Nvidia “NDR” Infiniband network
- 400 Gbps network bandwidth for CPU nodes
- 800 Gbps network bandwidth for GPU nodes
- Fully non-blocking, meaning every node can talk to every other node at full bandwidth simultaneously.
- Parallel storage: 29 petabytes, NVMe SSD based storage from VAST Data.