← Back to Mini PC NVIDIA

NVIDIA DGX Spark

AI Development, Machine Learning, Large Language Models, Data Science, Local AI Workstation

Overview

The NVIDIA DGX Spark is a compact personal AI supercomputer built on the NVIDIA GB10 Grace Blackwell Superchip. It pairs a 20-core Arm CPU (10x Cortex-X925 + 10x Cortex-A725) with a Blackwell GPU delivering up to 1 PFLOP of FP4 tensor performance, 128GB of coherent unified memory, and 4TB of self-encrypting NVMe storage. With ConnectX-7 200 Gbps networking, it is designed to run large AI and machine-learning models locally in a 150mm cube chassis.

Key Highlights

  • NVIDIA GB10 Grace Blackwell Superchip combining a 20-core Arm CPU with Blackwell GPU in one unified design
  • 128GB of coherent unified LPDDR5x system memory (273 GB/s) for large AI and machine-learning workloads
  • Up to 1 PFLOP of FP4 tensor performance (5th-gen Tensor Cores, 4th-gen RT Cores) for the DGX platform
  • ConnectX-7 200 Gbps networking plus 10GbE wired Ethernet and Wi-Fi 7 / Bluetooth 5.4 wireless
  • Compact 150mm cube chassis (1.2 kg) with 240W power supply (140W GB10 TDP) and AI-optimized DGX OS

Video

Technical Specifications

Brand NVIDIA
Model NVIDIA DGX Spark
Processor (CPU) NVIDIA GB10 Grace Blackwell Superchip, 20-core Arm (10x Cortex-X925 + 10x Cortex-A725)
Graphics (GPU) NVIDIA Blackwell Architecture (5th-gen Tensor Cores, 4th-gen RT Cores, up to 1 PFLOP FP4)
Memory (RAM) 128GB LPDDR5x Coherent Unified System Memory, 256-bit interface, 273 GB/s bandwidth
Storage 4TB NVMe M.2 (with self-encryption)
Ports & I/O 4x USB Type-C 1x HDMI 2.1a up to 3x DisplayPort over USB-C (DP Alt Mode) 1x RJ-45 10GbE HDMI multichannel audio output
Connectivity Wi-Fi 7, Bluetooth 5.4, ConnectX-7 NIC @ 200 Gbps
Network Wi-Fi 7, Bluetooth 5.4, ConnectX-7 200 Gbps NIC, 10GbE RJ-45
Operating System NVIDIA DGX OS
Dimensions 150 x 150 x 50.5 mm (1.2 kg)