NVIDIA has launched Project DIGITS, a groundbreaking personal AI supercomputer that gives researchers, data scientists, and students access to the immense capabilities of the new NVIDIA Grace Blackwell platform. The centerpiece of this initiative is the innovative NVIDIA GB10 Grace Blackwell Superchip, which offers a petaflop of AI computing performance, supporting the prototyping and execution of large AI models.
The Project DIGITS system enables users to develop and run their models directly on their desktops and facilitates smooth deployment on accelerated cloud or data center infrastructures. Jensen Huang, NVIDIA’s founder and CEO, emphasized the importance of placing an AI supercomputer on the desks of every AI researcher, data scientist, and student, empowering them to shape the future of AI while making advanced technology more accessible.
The GB10 Superchip is a sophisticated system-on-a-chip (SoC) that boasts up to 1 petaflop of AI performance at FP4 precision. It combines an NVIDIA Blackwell GPU with state-of-the-art CUDA cores, connected to a high-performance NVIDIA Grace CPU, which has 20 energy-efficient cores built on the Arm architecture. This design, developed in collaboration with MediaTek, is not only powerful but also energy-efficient, capable of running demanding processes through a standard electrical outlet.
Each Project DIGITS unit comes equipped with 128GB of unified memory and up to 4TB of NVMe storage, allowing developers to work with up to 200 billion-parameter language models, thus enhancing AI innovation. Two Project DIGITS systems can be connected to handle models with up to 405 billion parameters, significantly amplifying computational capacity.
By leveraging the Grace Blackwell architecture, developers can prototype AI models locally on their Project DIGITS hardware running Linux-based NVIDIA DGX OS and seamlessly transition them to NVIDIA DGX Cloud or other data center services. This flexibility enables the scaling of projects while maintaining a familiar environment.
Users will also have access to a wide array of NVIDIA’s AI software tools, facilitating model fine-tuning, data science acceleration, and various other applications through popular frameworks like PyTorch, Python, and Jupyter notebooks. For those looking to create AI applications, NVIDIA Blueprints and NIM microservices offer resources for research and development.
Project DIGITS is set to become available in May, with a starting price of $3,000. Interested customers are encouraged to sign up for notifications.
In summary, NVIDIA’s Project DIGITS democratizes access to AI supercomputing power, fostering creativity and innovation in various fields. This initiative not only empowers individuals with cutting-edge technology but also positions AI as a transformative force across industries, signaling a bright future for advancements in artificial intelligence.