The NVIDIA® DGX Spark™ is a personal AI computer designed to develop, test and run advanced models. Powered by the Grace Blackwell GB10 Superchip, a Blackwell GPU with 5th-generation Tensor Cores, 128 GB of unified memory and 4 TB NVMe storage, it delivers up to 1 PFLOP for local prototyping of models with up to 200 billion parameters.
Product customisation
Don't forget to save your customization to be able to add to cartarrow_drop_downRécapitulatif
Total :
The NVIDIA® DGX Spark™ is a personal computer designed specifically to meet the growing demands of AI development. Compact and powerful, it allows developers, researchers and innovation teams to work locally on complex models without constantly relying on cloud infrastructure. Its full integration within the NVIDIA software ecosystem makes it an ideal tool for prototyping, testing and refining AI projects directly from a workstation.
At the centre of the DGX Spark™ is the NVIDIA GB10 Grace Blackwell Superchip, combining a 20-core Arm CPU (10 Cortex-X925 + 10 Cortex-A725) with a Blackwell GPU equipped with 5th-generation Tensor Cores. This unified architecture delivers up to 1 petaFLOP of AI performance in FP4, enabling local experimentation with models reaching up to 200 billion parameters. Two systems can also be interconnected via NVIDIA ConnectX™ to handle even larger models.
The DGX Spark™ features 128 GB of unified LPDDR5x memory, essential for handling large datasets and ensuring optimal coherence between CPU and GPU thanks to NVLink™-C2C. Its 4 TB NVMe M.2 SSD provides high transfer speeds and enough space to store models, datasets and development environments.
Designed for seamless integration into professional environments, the DGX Spark™ offers comprehensive connectivity: USB Type-C, 1 GbE and 10 GbE Ethernet, Wi-Fi 7 and Bluetooth® 5.3 LE. Display output is provided via an HDMI™ 2.1a port with multichannel audio support. The system runs on NVIDIA DGX OS, based on Ubuntu Linux, with direct access to NVIDIA AI frameworks and tools such as CUDA, TensorRT, PyTorch, Jupyter and AI Workbench.
With dimensions of only 150 × 150 × 50.5 mm and a weight of 1.2 kg, the DGX Spark™ delivers exceptional power in a highly compact enclosure. It can be placed on a desk or in a shared workspace while maintaining the performance required for advanced AI projects.
The NVIDIA® DGX Spark™ represents a new generation of AI-dedicated personal workstations. Combining power, compactness and software integration, it enables users to explore, test and optimise sophisticated AI models while ensuring a smooth transition to production environments or cloud infrastructures.
|
Processor |
NVIDIA® Grace CPU – 20 cores (10 Cortex-X925 + 10 Cortex-A725) |
|
GPU |
NVIDIA® Blackwell – 5th-gen Tensor Cores – up to 1 PFLOP (FP4) |
|
Architecture |
Grace Blackwell GB10 Superchip – unified CPU + GPU |
|
Memory |
128 GB unified LPDDR5x |
|
Storage |
4 TB NVMe M.2 self-encrypting SSD |
|
Wired Network |
1 x RJ45 (1 GbE), 1 x RJ45 (10 GbE), ConnectX-7 Smart NIC |
|
Wireless |
Wi-Fi 7, Bluetooth® 5.3 LE |
|
Audio |
Multichannel audio output via HDMI™ |
|
Rear I/O Ports |
1 x HDMI™ 2.1a, 4 x USB Type-C |
|
Encode / Decode |
1 × NVENC / 1 × NVDEC |
|
Operating System |
NVIDIA DGX™ OS (Ubuntu Linux) |
|
Dimensions & Weight |
150 × 150 × 50.5 mm — 1.2 kg |
|
Warranty |
2 years |
Data sheet