Sophia node architecture
Sophia is operated by Argonne Leadership Computing Facility in United States. This profile covers Documented compute profile.
Cluster interconnect: InfiniBand HDR200 · 8 compute ports per node. Topology: Fat-tree.
DGX A100 · 40 GB
- Node count
- 22
- CPU
- AMD Rome · 64 cores
- CPUs per node
- 2
- Accelerator
- NVIDIA A100
- Physical accelerators per node
- 8
- Host / unified memory
- 1 TB DDR4
- Accelerator memory
- 40 GB per GPU
- Local accelerator links
- NVLink / NVSwitch
- CPU-to-accelerator link
- Not verified
- Storage
- 4 × 3.84 TB NVMe
DGX A100 · 80 GB
- Node count
- 2
- CPU
- AMD Rome · 64 cores
- CPUs per node
- 2
- Accelerator
- NVIDIA A100
- Physical accelerators per node
- 8
- Host / unified memory
- 2 TB DDR4
- Accelerator memory
- 80 GB per GPU
- Local accelerator links
- NVLink / NVSwitch
- CPU-to-accelerator link
- Not verified
- Storage
- 4 × 3.84 TB NVMe
Scope and uncertainty
Twenty-two 320 GB GPU-memory nodes and two 640 GB GPU-memory nodes. Memory capacities are kept separate by node type.
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