Delta

NCSA · University of Illinois

Node type

CPU Accelerator Interconnect

Reading this map

Delta node architecture

Delta is operated by NCSA · University of Illinois in United States. This profile covers the documented compute inventory. Specifications are listed below; anything the operator has not published is itemised in the documented-fields list.

Cluster · A100 ×4

Operator / site: NCSA · University of Illinois
Country: United States
Profile scope: Documented compute inventory
Data checked: 13 September 2026
Operating system: Red Hat Enterprise Linux 9

A100 ×4

Compute nodes of this type: 100
CPU model: AMD EPYC 7763 · 64 cores
CPU sockets per node: 1
Cores per socket: 64
Physical CPU cores: 64
Accelerator model: NVIDIA A100
Physical accelerators per node: 4
Host / unified memory: 256 GB
Accelerator memory: 40 GB per GPU
Accelerator ↔ accelerator link: NVLink · all GPU pairs
On-node fabric: NVLink — four bonded links between every GPU pair (NV4 in the operator affinity matrix)
CPU ↔ accelerator link: PCIe 4.0
Interface speed: 200 Gb/s (HPE Slingshot 11)
Local storage: 1.5 TB local

A40 ×4

Compute nodes of this type: 100
CPU model: AMD EPYC 7763 · 64 cores
CPU sockets per node: 1
Cores per socket: 64
Physical CPU cores: 64
Accelerator model: NVIDIA A40
Physical accelerators per node: 4
Host / unified memory: 256 GB
Accelerator memory: 48 GB per GPU
Accelerator ↔ accelerator link: PCIe Gen4 through the host — no GPU-to-GPU NVLink
On-node fabric: None — NCSA's affinity matrix shows SYS (via the host bridge) between all four A40 GPUs
CPU ↔ accelerator link: PCIe Gen4
Interface speed: 200 Gb/s (HPE Slingshot 11)
Local storage: 1.5 TB local

CPU only

Compute nodes of this type: 132
CPU model: AMD EPYC 7763 · 64 cores
CPU sockets per node: 2
Cores per socket: 64
Physical CPU cores: 128
Host / unified memory: 256 GB
Interface speed: 200 Gb/s (HPE Slingshot 11)
Local storage: 0.74 TB local

A100 ×8

Compute nodes of this type: 6
CPU model: AMD EPYC 7763 · 64 cores
CPU sockets per node: 2
Cores per socket: 64
Physical CPU cores: 128
Accelerator model: NVIDIA A100
Physical accelerators per node: 8
Host / unified memory: 2,048 GB
Accelerator memory: 40 GB per GPU
Accelerator ↔ accelerator link: NVLink / NVSwitch
On-node fabric: NVLink — twelve bonded links between every GPU pair (NV12 in the operator affinity matrix)
CPU ↔ accelerator link: PCIe 4.0
Interface speed: 200 Gb/s (HPE Slingshot 11)
Local storage: 1.5 TB local

H200 ×8

Compute nodes of this type: 8
CPU model: Intel Xeon Platinum 8558 · 48 cores
CPU sockets per node: 2
Cores per socket: 48
Physical CPU cores: 96
Accelerator model: NVIDIA H200
Physical accelerators per node: 8
Host / unified memory: 2,048 GB
Accelerator memory: 141 GB per GPU
Accelerator ↔ accelerator link: NVLink · all GPU pairs
On-node fabric: NVLink — eighteen bonded links between every GPU pair (NV18 in the operator affinity matrix)
Interface speed: 200 Gb/s (HPE Slingshot 11)
Local storage: 2.0 TB local

MI100 / MI210

Compute nodes of this type: 1
CPU model: AMD Milan · 64 cores
CPU sockets per node: 2
Cores per socket: 64
Physical CPU cores: 128
Accelerator model: 8 × MI100 + 1 × MI210
Physical accelerators per node: 9
Host / unified memory: 2,048 GB
Accelerator memory: 32 GB per GPU (operator table)
Interface speed: 200 Gb/s (HPE Slingshot 11)
Local 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.

Documented fields: 75 / 128

Every field this catalog checks. Fields listed above have a value; the remainder are shown here. A field marked absent is not publicly available in the sources we could verify, which does not mean the hardware is missing. Fields that cannot apply, such as benchmark results for an unranked system, are excluded from the total.

Cluster — 1 / 3

Manufacturer
Ranked core count
Theoretical peak (PFlop/s)
Measured power (kW)
Energy efficiency (GFlop/s per watt)
Operating system
Network topology

A100 ×4 — 14 / 22

System / node model
Compute nodes of this type
CPU model
CPU sockets per node
Cores per socket
Physical CPU cores
CPU note
Accelerator model
Physical accelerators per node
Host / unified memory
Accelerator memory
Accelerator ↔ accelerator link
On-node fabric
CPU ↔ accelerator link
Network interfaces per node
Interface model
Interface speed
Interface-to-accelerator topology
Aggregate scale-out per node
GPUDirect RDMA
Management / storage networking
Local storage
NVMe data disks
Max remote data disks
Uncached disk IOPS
Uncached disk throughput (MBps)
Form factor

A40 ×4 — 14 / 22

System / node model
Compute nodes of this type
CPU model
CPU sockets per node
Cores per socket
Physical CPU cores
CPU note
Accelerator model
Physical accelerators per node
Host / unified memory
Accelerator memory
Accelerator ↔ accelerator link
On-node fabric
CPU ↔ accelerator link
Network interfaces per node
Interface model
Interface speed
Interface-to-accelerator topology
Aggregate scale-out per node
GPUDirect RDMA
Management / storage networking
Local storage
NVMe data disks
Max remote data disks
Uncached disk IOPS
Uncached disk throughput (MBps)
Form factor

CPU only — 8 / 15

System / node model
Compute nodes of this type
CPU model
CPU sockets per node
Cores per socket
Physical CPU cores
CPU note
Accelerator model
Physical accelerators per node
Host / unified memory
Accelerator memory
Accelerator ↔ accelerator link
On-node fabric
CPU ↔ accelerator link
Network interfaces per node
Interface model
Interface speed
Interface-to-accelerator topology
Aggregate scale-out per node
GPUDirect RDMA
Management / storage networking
Local storage
NVMe data disks
Max remote data disks
Uncached disk IOPS
Uncached disk throughput (MBps)
Form factor

A100 ×8 — 14 / 22

System / node model
Compute nodes of this type
CPU model
CPU sockets per node
Cores per socket
Physical CPU cores
CPU note
Accelerator model
Physical accelerators per node
Host / unified memory
Accelerator memory
Accelerator ↔ accelerator link
On-node fabric
CPU ↔ accelerator link
Network interfaces per node
Interface model
Interface speed
Interface-to-accelerator topology
Aggregate scale-out per node
GPUDirect RDMA
Management / storage networking
Local storage
NVMe data disks
Max remote data disks
Uncached disk IOPS
Uncached disk throughput (MBps)
Form factor

H200 ×8 — 13 / 22

System / node model
Compute nodes of this type
CPU model
CPU sockets per node
Cores per socket
Physical CPU cores
CPU note
Accelerator model
Physical accelerators per node
Host / unified memory
Accelerator memory
Accelerator ↔ accelerator link
On-node fabric
CPU ↔ accelerator link
Network interfaces per node
Interface model
Interface speed
Interface-to-accelerator topology
Aggregate scale-out per node
GPUDirect RDMA
Management / storage networking
Local storage
NVMe data disks
Max remote data disks
Uncached disk IOPS
Uncached disk throughput (MBps)
Form factor

MI100 / MI210 — 11 / 22

System / node model
Compute nodes of this type
CPU model
CPU sockets per node
Cores per socket
Physical CPU cores
CPU note
Accelerator model
Physical accelerators per node
Host / unified memory
Accelerator memory
Accelerator ↔ accelerator link
On-node fabric
CPU ↔ accelerator link
Network interfaces per node
Interface model
Interface speed
Interface-to-accelerator topology
Aggregate scale-out per node
GPUDirect RDMA
Management / storage networking
Local storage
NVMe data disks
Max remote data disks
Uncached disk IOPS
Uncached disk throughput (MBps)
Form factor

Sources

Browse all cluster architecture pages →