NVIDIA has introduced two groundbreaking AI supercomputers designed to bring data center-level performance to desktop environments. The DGX Spark and DGX Station aim to revolutionize AI development by providing researchers, data scientists, and students with powerful tools for training, running inference, and deploying large language models right at their desks.
DGX Spark: AI Supercomputing in a Mini PC Form Factor
The DGX Spark, previously known as Project Digits, is NVIDIA’s latest innovation in compact AI computing. This small yet mighty device packs an impressive punch, featuring the GB10 Grace Blackwell Superchip.
Key Specifications:
- Processor: NVIDIA GB10 Grace Blackwell Superchip
- AI Performance: 1,000 TOPS (Trillion Operations Per Second) for FP4 precision
- Memory: 128 GB LPDDR5x with 273 GB/s bandwidth
- Storage: 1 TB or 4 TB NVMe M.2 SSD
- Connectivity: 10GbE, WiFi 7, 4x USB4 ports
- Display Output: HDMI 2.1a
- Power Consumption: 170 Watts
- Dimensions: 150 x 150 x 50.5 mm
- Weight: 1.2 kg
The GB10 Superchip combines a 20-core Arm-based CPU (featuring 10 Cortex-X925 and 10 Cortex-A725 cores) with NVIDIA’s latest Blackwell GPU architecture. This integration allows for seamless CPU-GPU communication, crucial for complex AI workloads.
Advanced Interconnect Technology
A standout feature of the DGX Spark is its use of NVIDIA’s NVLink-C2C (Chip-to-Chip) interconnect. This technology provides 5 times the bandwidth of PCIe Gen 5, enabling rapid data transfer between the CPU and GPU. Such high-speed communication is essential for AI tasks that require frequent data movement between processing units.
Software Ecosystem
The DGX Spark runs on NVIDIA DGX OS, based on Ubuntu 22.04. It comes pre-loaded with NVIDIA’s AI software stack, including tools, frameworks, and pre-trained models. This ecosystem supports popular AI models from companies like DeepSeek, Meta, and Google, capable of handling models with up to 200 billion parameters locally.
For even larger models, users can link two DGX Spark systems together using NVIDIA ConnectX networking technology. This configuration allows for working with AI models containing up to 405 billion parameters.
Availability and Pricing
NVIDIA has announced that the DGX Spark will be available for pre-order at a price point of around $3,000. The device is expected to hit the market in May 2025, with manufacturing partners including ASUS, Dell, and HP.
DGX Station: Desktop Supercomputing for Advanced AI Development
For those requiring even more computational power, NVIDIA has unveiled the DGX Station. This desktop supercomputer is designed to bring data center-level performance to individual workstations.
Key Features:
- Processor: NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
- AI Performance: Up to 20 petaflops
- System Memory: 784 GB unified memory
- Networking: NVIDIA ConnectX-8 SuperNIC (up to 800 Gb/s)
The DGX Station leverages the GB300 Grace Blackwell Ultra Desktop Superchip, which integrates a high-performance NVIDIA Grace CPU with a Blackwell Ultra GPU. This combination delivers exceptional processing power for complex AI workloads.
Expandability and Collaboration
The DGX Station’s networking capabilities allow for connecting multiple units, enabling collaborative work on massive AI models. This feature is particularly useful for research teams or organizations working on cutting-edge AI projects that require immense computational resources.
Software Integration
Like its smaller counterpart, the DGX Station is compatible with NVIDIA’s comprehensive AI software stack. It supports the NVIDIA CUDA-X AI platform, NVIDIA NIM microservices, and the NVIDIA AI Enterprise software platform, providing a robust environment for AI development and deployment.
Availability
NVIDIA has announced that the DGX Station will be available later in 2025. Manufacturing partners for this high-end system include ASUS, BOXX, Dell, HP, Lambda, and Supermicro. Pricing details for the DGX Station have not been disclosed at the time of the announcement.
These new offerings from NVIDIA represent a significant step forward in making advanced AI computing accessible to a broader range of professionals and researchers. By bringing supercomputer-level performance to desktop form factors, NVIDIA aims to accelerate AI innovation across various fields, from academic research to industrial applications.
The introduction of the DGX Spark and DGX Station underscores the growing demand for powerful, localized AI computing solutions. As AI continues to permeate various industries, these compact supercomputers are poised to play a crucial role in driving the next wave of AI-powered innovations and discoveries.