NVIDIA DGX Spark, a personal supercomputer for AI work
NVIDIA DGX Spark is a new generation desktop AI computer built for work with large generative models without the need for a data centre. It delivers high compute performance, a large pool of memory and a professional software stack, which makes it a fit for developers, researchers, data analysts and everyone working with advanced AI technology.
Data centre performance on a desk
The system is built on the NVIDIA GB10 Grace Blackwell superchip, which combines a Blackwell GPU with fifth generation Tensor Cores and a 20-core NVIDIA Grace (Arm) processor.
This architecture delivers up to 1 PFLOPS of AI performance and noticeably speeds up data preparation, training and model inference.
Models up to 200 billion parameters
DGX Spark carries 128 GB of unified memory, which lets you run, test and fine-tune models of up to 200B parameters right at your desk.
Thanks to built-in NVIDIA ConnectX™ networking, two DGX Spark units can be linked to work with models of up to 405B parameters, including Llama 3.1 405B.
A development environment out of the box
The unit ships with NVIDIA DGX OS based on Ubuntu, a preconfigured NVIDIA AI stack and access to NVIDIA NIM™ and Blueprints.
Popular environments and frameworks are supported:
This makes it straightforward to prototype quickly and move workloads to a cloud or a data centre later.
Key features of DGX Spark
- NVIDIA GB10 Grace Blackwell superchip
- NVIDIA Blackwell GPU with 5th generation Tensor Cores
- 20-core NVIDIA Grace (Arm) processor
- Up to 1 PFLOPS of AI performance with FP4 support
- 128 GB of unified system memory
- Support for models up to 200B parameters
- Two units can be linked for models up to 405B parameters
- 4 TB self-encrypting NVMe M.2
- Compact desktop form factor
- Networking: 10 GbE, ConnectX-7 SmartNIC
- Wi-Fi 7 and Bluetooth 5.3 support
- 1× HDMI 2.1a, USB-C ports, NVENC / NVDEC
- OS: NVIDIA DGX OS
NVIDIA DGX Spark is a compact yet powerful system that puts data centre class compute on your desk. It suits work with advanced generative models, research projects and commercial AI deployments.
Choosing between a desktop unit and a cluster?
Tell us which models you plan to run and how many people will work with them. We will calculate whether one node is enough or a GPU server makes more sense.
We supply as an official PNY partner, with the manufacturer warranty.
Message us on Telegram or leave a request.
Read the review and measurements: NVIDIA DGX Spark: what it can really do — review →
