NVIDIA HGX Rubin NVL8

baseboard
8× Rubin
NVLink 6 Switch, 28.8 TB/s total
2026
FP8
140.0
PFLOPS
NVFP4
280.0
PFLOPS
Power
kW Total
Memory
2359
GB Total

Eight Rubin SXM on an HGX baseboard, listed by NVIDIA alongside HGX B300 and HGX B200. NVIDIA publishes NVFP4 twice and the two are different workloads, not a sparse/dense pair: inference 400 PFLOPS sparse, training 280 PFLOPS dense. FP8/FP6 training is 140 PFLOPS dense. System power is not published for the baseboard; the 24 kW figure on DGX Rubin NVL8 is the full chassis including CPUs. NVIDIA marks the specification preliminary.

We will point you at suppliers who have it. Free, and no signup.

FP8
140.00
PFLOPS
NVFP4
280.00
PFLOPS
FP6
140.00
PFLOPS

System Details

GPU Configuration

GPU Model: NVIDIA Rubin SXM
GPU Count: 8 GPUs
Architecture: Rubin
Interconnect: NVLink 6 Switch, 28.8 TB/s total

System Specifications

Form Factor: baseboard
Total Power:
Total Memory: 2359 GB
Memory Bandwidth: 176000 GB/s

Precision Performance Breakdown

PrecisionSystem PerformancePer GPUEfficiency
140.000 PFLOPS 17500.0 TFLOPS
280.000 PFLOPS 35000.0 TFLOPS
FP6 140.000 PFLOPS 17500.0 TFLOPS

All figures are dense. Vendors commonly headline the number, which is twice the dense one.

Powered by NVIDIA Rubin

This system utilizes 8 × NVIDIA Rubin SXM GPUs, each delivering exceptional performance for AI and HPC workloads.

Per GPU TDP

Per GPU Memory

288 GB

Process Node

Architecture

Rubin

Documentation & Resources

Official Datasheet

NVIDIA HGX Rubin NVL8 technical specifications

Download PDF

workstation-datasheet-dgx-spark-gtc25-spring-nvidia-us-3716899-web.pdf

NVIDIA • 2025-10-16

View Document ↗

NVIDIA DGX Documentation

System guides, deployment resources

Browse ↗

Typical Use Cases

Large Language Model Training
Distributed Deep Learning
Multi-GPU Inference
HPC Simulations
Scientific Computing

The NVIDIA HGX Rubin NVL8 runs 8× Rubin GPUs over NVLink 6 Switch, 28.8 TB/s total, delivering 140 PFLOPS FP8 dense.