Trillium is a cluster of 241,056 cores and 252 GPUs, owned by the University of Toronto and operated by SciNet. This system consists of a 1224-node CPU subcluster and a 63-node GPU subcluster. Trillium is intended to enable large parallel jobs of 1024 cores and more or requiring many GPUs. It is the one of the most powerful supercomputers in Canada available for academic research. Compute allocations are handled through the annual resource allocation competition of the Digital Research Alliance of Canada.
Trillium was officially launched on August 7, 2025, replacing the clusters Niagara and Mist.
Specifications of the cluster:
- Trillium’s CPU subcluster consists of 1,224 CPU compute node.
- The CPU compute nodes each have 196 AMD “Turin” cores at 2.6GHz, for a total of 235,008 cores.
- Trillium’s GPU subcluster consists of 63 GPU compute nodes
- The GPU compute nodes each have 96 AMD “Genoa” cores at 2.4 GHz and 4 NVIDIA H100 (80GB) GPUs, for a total of 252 GPUs.
- There is 768 GB of RAM per node.
- NDR InfiniBand 1:1 interconnect at 400 Gb/s for CPU nodes and 800 Gb/s for GPU nodes.
- 29 PB VAST NVMe storage for home, scratch, and project space.
- No local disks.
- Theoretical peak performance (“Rpeak”) of about 34 PF.
- About 1700 kW power consumption.