Isambard-AI phase 2 node architecture
Isambard-AI phase 2 is operated by University of Bristol in United Kingdom. This profile covers Ranked system / partition.
Cluster interconnect: Slingshot-11. Topology: Not verified.
GH200 compute
- Node count
- 1320
- CPU
- NVIDIA Grace · 72 cores
- CPUs per node
- 4
- Accelerator
- NVIDIA Hopper (GH200)
- Physical accelerators per node
- 4
- Host / unified memory
- 480 GB LPDDR5X · 460 GB usable
- Accelerator memory
- 96 GB per GPU
- Local accelerator links
- NVLink
- CPU-to-accelerator link
- NVLink-C2C
- Storage
- Not verified
Scope and uncertainty
Phase 2 only: 1,320 compute nodes. The 42 Phase 1 nodes and login nodes are excluded. Four 200 Gb/s Slingshot 11 adapters per compute node.
Not verified means a detail has not been confirmed in this catalog. Diagrams are schematic; node indices are illustrative, not physical hostnames. Data checked 14 September 2026.
Sources
Browse all cluster architecture pages →
CLUSTER