NVIDIA DGX Spark Founders Edition Personal AI Supercomputer

NVIDIA DGX Spark

NVIDIA DGX Spark is a personal AI computer designed from the ground up to build, fine-tune, and run AI models locally. Powered by the NVIDIA GB10 Grace Blackwell Superchip, it delivers up to 1 petaFLOP of AI performance while maintaining a compact desktop form factor suitable for developers, researchers, data scientists, and students.

Built for Large AI Workloads

As generative AI models continue to grow in size and complexity, local development requires significantly more memory and compute resources. NVIDIA DGX Spark addresses these demands by providing powerful AI acceleration and unified memory, enabling users to prototype, fine-tune, and run large AI models directly on their desktop.

Up to 200 Billion Parameter Models

Equipped with 128GB of coherent unified system memory and support for the FP4 data format, NVIDIA DGX Spark can support AI models with up to 200 billion parameters. This allows developers to experiment, fine-tune, and perform inference on advanced large language models without relying entirely on cloud infrastructure.

NVIDIA Grace Blackwell Superchip

At the core of NVIDIA DGX Spark is the NVIDIA GB10 Grace Blackwell Superchip, featuring an NVIDIA Blackwell GPU with fifth-generation Tensor Cores and FP4 support. The system also includes a high-performance 20-core NVIDIA Grace Arm CPU, enabling faster data preprocessing, orchestration, model tuning, and inference workloads.

High-Speed CPU and GPU Memory Architecture

The GB10 Superchip utilizes NVIDIA NVLink-C2C technology to create a coherent CPU and GPU memory model. This architecture delivers up to five times the bandwidth of PCIe Gen 5, allowing the CPU and GPU to efficiently share memory and accelerate AI workloads.

Scale to Larger Models

Using integrated NVIDIA ConnectX networking, two NVIDIA DGX Spark systems can be connected together to support inference on AI models with up to 405 billion parameters. This enables users to work with even larger foundation models while maintaining a local development environment.

Preconfigured AI Software Stack

NVIDIA DGX Spark comes preconfigured with NVIDIA DGX OS based on Ubuntu Linux and includes the latest NVIDIA AI software stack. Developers can immediately start working with popular tools and frameworks such as PyTorch, Jupyter, Ollama, CUDA, TensorRT, cuDNN, TensorFlow, MATLAB, and NVIDIA NIM services.

Develop Locally, Deploy Anywhere

The software architecture mirrors the same environment used in NVIDIA AI factories and enterprise AI infrastructure. This allows developers to seamlessly move projects from NVIDIA DGX Spark to DGX Cloud, accelerated cloud platforms, or data center environments with minimal workflow changes.

Compact Desktop AI Workstation

Despite its powerful AI capabilities, NVIDIA DGX Spark maintains a compact desktop design and includes 4TB of NVMe storage. The system combines high-performance AI computing, large memory capacity, and enterprise-grade software tools into a workstation suitable for local AI development and experimentation.