Nvidia has announced an exciting new product, Project Digits, a personal AI supercomputer that will be available for $3,000.
The system, which is expected to launch in May, features the powerful new GB10 Grace Blackwell Superchip, allowing it to run complex AI models while remaining compact enough to fit on a desk and run from a standard power outlet.
This is a significant leap forward in AI computing, as systems with this level of processing power would typically require larger, more power-hungry machines.
Project Digits is capable of handling AI models with up to 200 billion parameters and offers an impressive starting configuration with 128GB of unified memory, which is much more than what most laptops offer, along with up to 4TB of NVMe storage.
For those needing even more power, two Project Digits systems can be linked together to support models with up to 405 billion parameters, matching the scale of Meta’s Llama 3.1 AI model.
The GB10 chip at the heart of Project Digits delivers up to 1 petaflop of AI performance, which means it can perform a mind-boggling 1 quadrillion calculations per second.
The chip also utilizes Nvidia’s latest-generation CUDA cores and fifth-generation Tensor Cores, which are connected through NVLink-C2C to a Grace CPU containing 20 energy-efficient Arm-based cores.
Nvidia worked with MediaTek, a company known for its Arm-based chips, to optimize the power efficiency and performance of the chip, ensuring it can handle intense workloads while using less power than previous systems.
Along with the hardware, Nvidia provides users access to a rich software library, which includes development kits, orchestration tools, and pre-trained models available through the Nvidia NGC catalog.
The system runs on Nvidia DGX OS, a Linux-based operating system, and supports popular AI frameworks like PyTorch, Python, and Jupyter Notebooks.
