Nvidia shrinks Grace-Blackwell superchip for AI mini-PC
Marian Schultz
▶January 7, 2025
Become smarter in just 5 minutes
Feeeed serve up the most interesting stories in AI, tech, business, science, and more—all in 5 minutes or less.
Booyah! The first newsletter will go out early next week.
Oops! Something went wrong while submitting the form.
The Rundown: Nvidia unveiled Project Digits, a $3,000 mini-PC powered by the Grace-Blackwell GB10 superchip, featuring 128GB of memory and tailored for AI developers to run large models on the desktop.
The Context:
- The system pairs Nvidia's Arm-based Grace CPU with a Blackwell GPU, offering up to 500 TFLOPS performance on compressed 4-bit models.
- It supports models up to 200 billion parameters or 405 billion when linked with another unit via onboard ConnectX networking.
- Packed with 128GB LPDDR5x memory and 4TB NVMe storage, the device aims at AI inferencing, model fine-tuning, and edge applications.
- It will launch in May, echoing Nvidia’s push for desktop AI systems after years of Jetson dev kits.
Why it Matters: Nvidia’s Project Digits makes running large AI models accessible on compact hardware, signaling a new era of affordable, high-performance AI desktops. While it won’t dethrone existing workstations, it’s a leap for students, researchers, and developers working outside data centers.
