NVIDIA DGX Spark Reviews & Specifications
A Grace Blackwell AI Supercomputer on Your Desk
Discover the revolutionary NVIDIA DGX Spark - the world's first desktop AI supercomputer powered by Grace Blackwell architecture. Get comprehensive reviews, detailed specifications, and hands-on experiences from developers and researchers worldwide.
Latest DGX Spark Reviews & Videos
Watch the latest NVIDIA DGX Spark reviews and hands-on experiences from developers, researchers, and tech enthusiasts worldwide. Discover real-world performance, setup guides, and AI development workflows.

128GB with 200GbE NVIDIA DGX Spark is GREAT for Local AI- by ServeTheHome | DGX Spark Review | How about DGX Spark?
This ServeTheHome video explores the NVIDIA DGX Spark, a compact AI workstation. It boasts impressive specs, including 128GB of memory and 200GbE networking, within a surprisingly low power envelope. The review demonstrates setup, software, and scaling options for multiple units.
October 20, 2025

I got a desktop supercomputer? | NVIDIA DGX Spark overview
This video unboxes and reviews the NVIDIA DGX Spark, a dedicated AI workstation. The creator explores its capabilities, demonstrating GPT-OSS 120B and fine-tuning a Gemma 4B model for customer support. Expect a deep dive into the device's hardware and software, including its unique operating system.
October 17, 2025

Getting Started with DGX Spark - Setup Guide and First Steps
A complete guide to setting up your NVIDIA DGX Spark and running your first AI models. Learn how to configure the system, install additional software, and start developing.
March 18, 2025

NVIDIA DGX Spark – A Non-Sponsored Review (Strix Halo Comparison, Pros & Cons) | NVIDIA DGX Spark Review
This unsponsored NVIDIA DGX Spark review compares it to the AMD Strix Halo. The video details the unboxing, setup, and performance tests for video and image generation, as well as LLM inference. Direct comparisons across several models reveal potential advantages and disadvantages.
March 15, 2025

Unboxing the NVIDIA DGX Spark - First Impressions
Watch as we unbox the revolutionary NVIDIA DGX Spark and share our first impressions of this desktop AI supercomputer powered by the Grace Blackwell architecture.
March 15, 2025
About NVIDIA DGX Spark
Revolutionary Desktop AI Supercomputer Powered by Grace Blackwell
Breakthrough AI Performance
The DGX Spark delivers unprecedented AI performance in a compact desktop form factor. With up to 1 petaFLOP of AI computing power, it enables local development and inference of large language models with up to 200 billion parameters.
Unified Memory Architecture
Featuring 128 GB of LPDDR5x unified system memory with 273 GB/s bandwidth, the DGX Spark eliminates traditional memory bottlenecks between CPU and GPU, enabling seamless data sharing and larger model processing.
Pre-installed AI Software Stack
The DGX Spark comes with NVIDIA DGX OS pre-installed, including NVIDIA NIM, CUDA toolkit, and optimized AI frameworks. Get started immediately with popular models from DeepSeek, Meta, Google, Qwen, and others.
Why Choose DGX Spark
- Local AI development without cloud dependencies
- Complete data privacy and security
- Energy-efficient 240W power consumption
- Compact 150mm × 150mm × 50.5mm form factor
Ideal For
The DGX Spark is perfect for AI researchers, developers, data scientists, and educational institutions.
AI model research and experimentation
Prototyping and fine-tuning AI models
Large-scale data processing and analysis
AI/ML learning and academic projects
DGX Spark Performance Metrics
DGX Spark Key Features
NVIDIA GPU, CPU, Networking, and AI Software Technologies
NVIDIA GB10 Grace Blackwell Superchip
Experience up to 1 petaFLOP of AI performance at FP4 precision with the revolutionary NVIDIA Grace Blackwell architecture, featuring 20-core ARM CPU and Blackwell GPU.
128 GB LPDDR5x Unified System Memory
Run AI development and testing workloads with models up to 200 billion parameters. The unified memory architecture provides 273 GB/s bandwidth for seamless data sharing between CPU and GPU.
NVIDIA ConnectX-7 Networking
High-performance 200 Gbps NVIDIA ConnectX-7 networking enables connection of two DGX Spark systems to work with AI models of up to 405 billion parameters.
NVIDIA AI Software Stack
Pre-installed NVIDIA DGX OS with full AI software stack including NVIDIA NIM, CUDA toolkit, and optimized frameworks for generative AI workloads.
DGX Spark Technical Specifications
Architecture | NVIDIA Grace Blackwell Architecture |
Superchip | NVIDIA GB10 Grace Blackwell Superchip |
CPU Cores | 20-core ARM: 10 Cortex-X925 + 10 Cortex-A725 |
GPU | Blackwell Architecture with 5th Gen Tensor Cores |
System Memory | 128 GB LPDDR5x Unified System Memory |
Memory Bandwidth | 273 GB/s Memory Bandwidth |
AI Performance | Up to 1 PetaFLOP (FP4) AI Performance |
Storage | 4 TB NVMe M.2 with Self-Encryption |
Networking | NVIDIA ConnectX-7 NIC @ 200 Gbps |
Connectivity | WiFi 7, Bluetooth 5.4, 4x USB-C, 1x HDMI 2.1a |
Power Supply | 240 Watts Power Supply |
Dimensions & Weight | 150mm × 150mm × 50.5mm (1.2kg) |
Operating System | NVIDIA DGX OS Pre-installed |
DGX Spark Frequently Asked Questions
Get answers to common questions about NVIDIA DGX Spark setup, usage, and troubleshooting
Ready to Experience Desktop AI Supercomputing?
Get your NVIDIA DGX Spark today and start developing with the power of Grace Blackwell architecture.