Next upHack for Humanity: San Francisco (powered by Google Gemini)
News · ces

Nvidia at CES 2025: a desktop supercomputer that can run 200B parameter LLMs

Announced this Monday at CES, Nvidia's Project DIGITS personal AI supercomputer aims to put a petaflop of AI processing power in the hands of every data scientist, AI researcher and student. The small and mighty unit packs a GB10 Grace Blackwell Superchip and can run models of up to 200 parameters.

Dmytro Spodarets
Jan 9, 2025 · 1 min read

On Monday at CES, Nvidia unveiled Project DIGITS, a $3000 personal AI supercomputer sporting the novel NVIDIA GB10 Grace Blackwell Superchip to deliver a petaflop of computing performance, enough for prototyping, fine-tuning and running LLMs up to 200 billion parameters in size. In addition to the GB10, Project DIGITS units feature Wi-Fi, Bluetooth and USB connectivity, 128 GB of unified memory and a 4 TB solid-state drive. Nvidia's ConnectX technology enables two Project DIGITS devices to be linked together to double up capacity and run LLMs with as many as 405 billion parameters.

The new NVIDIA GB10 Grace Blackwell Superchip is a system-on-a-chip (SoC) featuring an Nvidia Blackwell GPU with latest-generation CUDA cores and fifth-generation Tensor Cores connected to a 20-Arm core Nvidia Grace CPU. Nvidia built the SoC in collaboration with leading Arm SoC designer MediaTek. Project DIGITS harnesses the power of the GB1o through Nvidia's Linux-based DGX operating system. Additionally, and delivering Nvidia CEO Jensen Huang's promise that Project DIGITS runs the entire Nvidia AI stack, Project DIGITS have access to Nvidia's software and libraries, including the NVIDIA NeMo framework for model fine-tuning, RAPIDS libraries for data science, and NVIDIA Blueprints and NVIDIA NIM microservices for agentic AI applications.

Project DIGITS is expected to be available this May.


Dmytro Spodarets
Dmytro Spodarets
Founder & Editor-in-Chief

Founder and Chief Editor of Data Phoenix — a San Francisco Bay Area media and education platform focused on AI and Data.

More news