Delta node architecture
Delta is operated by NCSA · University of Illinois in United States. This profile covers the documented compute inventory.
Cluster interconnect: HPE Slingshot 11 · 200 Gb/s. Topology: Not verified.
A100 ×4
- Node count
- 100
- CPU
- AMD EPYC 7763 · 64 cores
- CPUs per node
- 1
- Accelerator
- NVIDIA A100
- Physical accelerators per node
- 4
- Host / unified memory
- 256 GB
- Accelerator memory
- 40 GB per GPU
- Local accelerator links
- NVLink · all GPU pairs
- CPU-to-accelerator link
- PCIe 4.0
- Storage
- 1.5 TB local
A40 ×4
- Node count
- 100
- CPU
- AMD EPYC 7763 · 64 cores
- CPUs per node
- 1
- Accelerator
- NVIDIA A40
- Physical accelerators per node
- 4
- Host / unified memory
- 256 GB
- Accelerator memory
- 48 GB per GPU
- Local accelerator links
- PCIe via host; no direct NVLink in operator map
- CPU-to-accelerator link
- PCIe 4.0
- Storage
- 1.5 TB local
CPU only
- Node count
- 132
- CPU
- AMD EPYC 7763 · 64 cores
- CPUs per node
- 2
- Accelerator
- No accelerator
- Physical accelerators per node
- 0
- Host / unified memory
- 256 GB
- Accelerator memory
- Not verified
- Local accelerator links
- Not applicable
- CPU-to-accelerator link
- PCIe 4.0
- Storage
- 0.74 TB local
A100 ×8
- Node count
- 6
- CPU
- AMD EPYC 7763 · 64 cores
- CPUs per node
- 2
- Accelerator
- NVIDIA A100
- Physical accelerators per node
- 8
- Host / unified memory
- 2,048 GB
- Accelerator memory
- 40 GB per GPU
- Local accelerator links
- NVLink / NVSwitch
- CPU-to-accelerator link
- PCIe 4.0
- Storage
- 1.5 TB local
H200 ×8
- Node count
- 8
- CPU
- Intel Xeon Platinum 8558 · 48 cores
- CPUs per node
- 2
- Accelerator
- NVIDIA H200
- Physical accelerators per node
- 8
- Host / unified memory
- 2,048 GB
- Accelerator memory
- 141 GB per GPU
- Local accelerator links
- Not verified
- CPU-to-accelerator link
- Not verified
- Storage
- Not verified
MI100 / MI210
- Node count
- 1
- CPU
- AMD Milan · 64 cores
- CPUs per node
- 2
- Accelerator
- 8 × MI100 + 1 × MI210
- Physical accelerators per node
- 9
- Host / unified memory
- 2,048 GB
- Accelerator memory
- 32 GB per GPU (operator table)
- Local accelerator links
- Not verified
- CPU-to-accelerator link
- Not verified
- Storage
- 1.5 TB local
Scope and uncertainty
Compute-node counts follow individual specification tables. The operator page's introductory large-memory count is inconsistent with those tables.
Not verified means a detail has not been confirmed in this catalog. Diagrams are schematic; node indices are illustrative, not physical hostnames. Data checked 13 September 2026.
Sources
Browse all cluster architecture pages →
CLUSTER