Nvidia at present unveiled Nvidia Undertaking Digits, a private AI supercomputer that gives AI researchers, knowledge scientists and college students worldwide with entry to the ability of the Nvidia Grace Blackwellplatform.
Undertaking Digits options the brand new Nvidia GB10 Grace Blackwell Superchip, providing a petaflop of AI computing efficiency for prototyping, fine-tuning and operating giant AI fashions. The corporate made the announcement throughout CEO Jensen Huang’s opening keynote at CES 2025, the large tech commerce present in Las Vegas this week.
With Undertaking Digits, customers can develop and run inference on fashions utilizing their very own desktop system, then seamlessly deploy the fashions on accelerated cloud or knowledge middle infrastructure. It’s based mostly on a “super secret chip called GB110, the smallest Blackwell we can make,” Huang stated.
Nvidia Undertaking Digits
“AI will be mainstream in every application for every industry. With Project Digits, the Grace Blackwell Superchip comes to millions of developers,” stated Huang. “Placing an AI supercomputer on the desks of every data scientist, AI researcher and student empowers them to engage and shape the age of AI.”
GB10 Superchip offers a petaflop of power-efficient AI efficiency
The GB10 Superchip is a system-on-a-chip (SoC) based mostly on the Nvidia Grace Blackwell structure. It delivers as much as one petaflop of AI efficiency at FP4 precision.
GB10 options an Nvidia Blackwell GPU with latest-generation CUDA cores and fifth-generation Tensor Cores, related by way of NVLink-C2C chip-to-chip interconnect to a high-performance Nvidia Grace CPU, which incorporates 20 power-efficient cores constructed with the Arm structure. MediaTek, a market chief in Arm-based SoC designs, collaborated on the design of GB10, contributing to its best-in-class energy effectivity, efficiency and connectivity.
The GB10 Superchip permits Undertaking Digits to ship highly effective efficiency utilizing solely a typical electrical outlet. Every Undertaking Digits options 128GB of unified, coherent reminiscence and as much as 4TB of NVMe storage. With the supercomputer, builders can run as much as 200-billion-parameter giant language fashions. As well as, utilizing Nvidia ConnectX networking, two Undertaking Digits AI supercomputers will be linked to run as much as 405-billion-parameter fashions.
Grace Blackwell AI supercomputing inside attain
There are 72 Blackwell chips on this wafer.
With the Grace Blackwell structure, enterprises and researchers can prototype, fine-tune and take a look at fashions on native Undertaking Digits programs operating Linux-based Nvidia DGX OS, after which deploy them seamlessly on Nvidia DGX Cloud, accelerated cloud situations, or knowledge middle infrastructure.
This enables builders to prototype AI on Undertaking Digits after which scale on cloud or knowledge middle infrastructure, utilizing the identical Grace Blackwell structure and the Nvidia AI Enterprise software program platform.
Undertaking Digits customers can entry an in depth library of Nvidia AI software program for experimentation and prototyping, together with software program improvement kits, orchestration instruments, frameworks and fashions out there within the Nvidia NGC catalog and on the Nvidia Developer portal. Builders can fine-tune fashions with theNvidia NeMo framework, speed up knowledge science with Nvidia Rapids libraries and run frequent frameworks similar to PyTorch, Python and Jupyter notebooks.
To construct agentic AI purposes, customers can even harness Nvidia blueprints and Nvidia NIM microservices, which can be found for analysis, improvement and testing by way of the Nvidia Developer Program. When AI purposes are prepared to maneuver from experimentation to manufacturing environments, the Nvidia AI Enterprise license offers enterprise-grade safety, help and product releases of Nvidia AI software program.
Availability
Undertaking Digits is coming.
Undertaking Digits can be out there in Could from Nvidia and high companions, beginning at $3,000.
“By making Llama models open-source, we’re committed to democratizing access to cutting-edge AI technology. With Project Digits, developers can harness the power of Llama locally, unlocking new possibilities for innovation and collaboration,” stated Ahmad Al-Dahle, head of GenAI at Meta, in an announcement.
“Advancing AI requires tools that empower researchers to experiment at scale, speed and precision. Nvidia’s Project Digits represents a significant leap forward. I’m excited to see how 128GB in such a small form factor can advance the future of enterprise AI,” stated Silvio Savarese, chief scientist at Salesforce, in an announcement.
“At Hugging Face, we want to make it easy for developers to build their own AI,” stated Je Boudier, head of product at Hugging Face, in an announcement. “Nvidia’s Project Digits will empower AI builders to build and run their own Gen AI models and systems at the edge. With 128GB of unified memory, AI builders can run 200B parameter models locally, and connect multiple Project Digits systems to scale from there. I can’t wait to see what the Hugging Face community will build with Nvidia Project Digits.”
“Nvidia’s Project Digits is a powerhouse you can hold in the palm of your hand. With two Project Digits units, developers can easily work with AI models up to 405B parameters in size. We can’t wait to see the apps people will build with this,” stated Michael Chiang, cofounder of Ollama, in an announcement.
Each day insights on enterprise use circumstances with VB Each day
If you wish to impress your boss, VB Each day has you coated. We provide the inside scoop on what corporations are doing with generative AI, from regulatory shifts to sensible deployments, so you’ll be able to share insights for max ROI.
An error occured.
