NVIDIA Announces DGX Spark and DGX Station
Personal AI Computers
Personal AI Computers
NVIDIA REVOLUTIONISES AI DEVELOPMENT WITH NEW DGX PERSONAL AI SUPERCOMPUTERS
NVIDIA is setting a new milestone for AI developers, researchers and students with the introduction of the NVIDIA DGX Spark and DGX Station. These powerful desktop supercomputers bring the Grace Blackwell architecture - previously only available in data centres - directly to the desktop.
DGX Spark - the world's smallest AI supercomputer - enables prototyping, fine-tuning and inference of large models locally or in the cloud. With the GB10 Grace Blackwell superchip, it offers up to 1,000 trillion operations per second and is perfect for generative AI and robotics.
DGX Station sets new standards with the GB300 Grace Blackwell Ultra Superchip: 784GB of coherent memory and ultra-fast 800Gb/s networking technology make it the most powerful desktop system for AI training and inference.
‘AI requires a new class of computing - and these DGX systems make it accessible,’ said NVIDIA CEO Jensen Huang.
Availability:
- DGX Spark is available for pre-order now.
- DGX Station coming later in 2025
Project DIGITS brings the power of a data centre to the desktop - ready to unleash the next era of AI development. With a compact design, advanced hardware and the ability to run all of NVIDIA's AI software, DIGITS sets a new benchmark for AI supercomputers. Goal: Create an affordable and scalable platform that accelerates the integration of AI into a wide range of industries and applications.
Features:
- Fully integrated platform: Runs with all NVIDIA AI software, including DGX Cloud.
- Flexibility: Can be used as a cloud computing platform, Linux workstation or wireless solution.
- Miniaturisation: Based on the new GB110 chip, the smallest Grace Blackwell chip from NVIDIA.
- Plug-and-play AI supercomputer: A complete solution that works right out of the box.
- Scalability: Ability to connect multiple devices via ConnectX and Nickel GPU Direct.
- Accessibility: For any user - PC, Mac or other systems.
The keynote focused on the launch of the NVIDIA GeForce RTX 5090 GPU, which sets new standards with 92 billion transistors and a performance of 3,352 trillion AI operations per second. The new GPUs, which will be available from 30 January, promise revolutionary advances in graphics with features such as RTX Neural Shaders, which improve lighting, textures and materials in real time.
In addition, DLSS 4, a new generation of Deep Learning Super Sampling technology, brings up to three additional frames per computed frame, greatly increasing efficiency.