GPU Architecture
NVIDIA Blackwell Ultra GPU

CPU
NVIDIA Grace CPU

Superchip
NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip

Tensor Cores
Fifth-Generation Tensor Cores (FP4 support)

Memory Capacity
775 GB Large Coherent Memory

Interconnect
NVIDIA NVLink-C2C Chip-to-Chip Interconnect

Networking
NVIDIA ConnectX-8 SuperNIC

Network Speed
Up to 800 Gb/s

Operating System
NVIDIA DGX OS

AI Software Platform
NVIDIA CUDA-X AI Platform

Workload Support
AI Training, AI Inference, Data Science, Machine Learning

Deployment
Local Workstation, DGX Cloud, or Data Center Infrastructure

System Capability
Workload-Optimized Power Shifting