NVIDIA Drops Project DIGITS: Personal AI Supercomputers for Researchers and Developers

NVIDIA just revealed Project DIGITS, a personal AI supercomputer that makes high-performance AI accessible to researchers, developers, and students worldwide. It’s built on the NVIDIA Grace Blackwell platform, offering a petaflop of AI power for training and running large-scale models.

The core of Project DIGITS is the new NVIDIA GB10 Grace Blackwell Superchip. This chip delivers a petaflop of AI performance, letting users fine-tune, prototype, and deploy AI models right from their desktops. Later, they can scale their work to cloud or data center environments without any extra hassle.

Key Features of Project DIGITS

• GB10 Grace Blackwell Superchip: Combines an NVIDIA Blackwell GPU with CUDA and Tensor Cores for AI and connects to an Arm-based NVIDIA Grace CPU.
• Massive Memory and Storage: Each system includes 128GB of unified memory and up to 4TB of NVMe storage.
• Energy Efficiency: Runs on a standard power outlet.
• Scalability: Two units can be linked via NVIDIA ConnectX networking to tackle larger models, up to 405 billion parameters.

With Grace Blackwell, users can prototype on their local systems and scale effortlessly using NVIDIA DGX Cloud or data centers. The project also integrates seamlessly with NVIDIA’s AI tools like NeMo, RAPIDS, and popular frameworks like PyTorch and Python.

Project DIGITS will hit the market in May, starting at $3,000.

Source: https://nvidianews.nvidia.com/news/nvidia-puts-grace-blackwell-on-every-desk-and-at-every-ai-developers-fingertips