NVIDIA DGX™ Rubin NVL8
NVIDIA DGX Rubin NVL8 is a turnkey infrastructure solution purpose-built to accelerate AI workloads and deliver intelligence at scale. Built on the NVIDIA Rubin architecture, it delivers high-performance compute, large GPU memory bandwidth, and sixth-generation NVIDIA NVLink connectivity for agentic AI, reasoning, training, and inference workloads.
- 8 NVIDIA Rubin GPUs
- 2.3 TB total GPU memory with 176 TB/s memory bandwidth
- Up to 280 petaFLOPS of NVFP4 training performance
- Up to 400 petaFLOPS of NVFP4 inference performance
DGX Rubin Technical Specifications
| Specification | Details |
|---|---|
| GPU and CPU |
|
| Total GPU Memory |
|
| GPU Memory Bandwidth |
|
| Performance |
|
| CPU |
|
| NVIDIA NVLink Bandwidth |
|
| System Power Usage |
|
| Networking |
|
| Management Network |
|
| Storage |
|
| Software |
|
| Rack Units |
|
| Enterprise Support |
|
** Dense specification
