NVIDIA recently introduced Project DIGITS, a personal AI supercomputer powered by the GB10 Superchip. Built on the Grace Blackwell architecture, it delivers high-performance computing for prototyping, fine-tuning, and running large AI models on a desktop.

The GB10 Superchip combines the NVIDIA Grace CPU and Blackwell GPU, offering up to one petaflop of AI performance at FP4 precision. It incorporates fifth-generation Tensor Cores and NVLink-C2C interconnect technology, emphasizing power efficiency and robust performance. Each Project DIGITS system is equipped with 128GB of unified memory and up to 4TB of NVMe storage, enabling it to handle models with up to 200 billion parameters. For larger-scale needs, systems can be interconnected to support even greater computational requirements.

Project DIGITS supports scaling from local development to deployment on NVIDIA DGX Cloud or data center infrastructure, maintaining consistency with the Grace Blackwell architecture.

 

The system is accompanied by NVIDIA AI Enterprise software, which includes a wide range of tools and frameworks to facilitate experimentation and development. Supported applications include popular frameworks such as PyTorch and NVIDIA NeMo for model development, along with RAPIDS libraries for data science workflows.

Further information

The system is expected to be available in May, with pricing starting at $3,000. Users can sign up for notifications on the NVIDIA website.