Join top executives in San Francisco on July 11-12 to hear how leaders are integrating and optimizing AI investments for success. Learn more
At this year’s GPU Technology Conference (GTC), Nvidia continued its AI hardware push with a specific focus on making its technology more accessible to enterprises across industries and streamlining the development of generative AI applications like ChatGPT.
The following is a daily recap of major announcements the Santa Clara, California-based company made with links to in-depth coverage.
Rent AI supercomputing infrastructure with DGX Cloud
While Nvidia has been building hardware for AI for quite some time now, it has taken some time for the technology to be widely adopted, in part due to its high cost. In 2020, the DGX A100 server box sold for $199,000. To change this, the company today announced DGX Cloud, a service that allows businesses to access its AI supercomputing infrastructure and software through a web browser. It rents out DGX Server boxes, each with eight Nvidia H100 or A100 GPUs and 640 GB of memory, and costs $36,999 per month for a single node.
Leveraging the power of DGX Cloud, the company also announced the launch of AI Foundations to help enterprises create and use custom generative AI models. The offering, Nvidia said, offers three cloud services: Nvidia NeMo for large language models (LLMs), Nvidia Picasso for image, video and 3D applications, and BioNeMO generate scientific texts based on biological data.
New hardware for AI inference and recommendations
In addition to DGX and AI Foundations, Nvidia also debuted four inference platforms designed to help developers quickly build specialized generative AI applications. This includes Nvidia L4 for producing AI video; Nvidia L40 for generating 2D/3D images; Nvidia H100 NVL for deploying large language models; and Nvidia Grace Hopper – connecting the Grace CPU and Hopper GPU through a super-fast 900GB/s coherent chip-to-chip interface – for recommendation systems built on massive datasets.
The company says L4 can deliver 120x more AI-powered video performance than CPUs, combined with 99% better power efficiency; while L40 acts as the engine of Omniverse, delivering 7x the inference performance for stable diffusion and 12x Omniverse performance over the previous generation.
Chip makers get cuLitho with Nvidia GTC
During the event, Nvidia CEO Jenson Huang took the stage to announce Nvidia cuLitho software library for computational lithography. The offering, as Huang explained, enables semiconductor companies to design and develop chips with ultra-small transistors and wires, while accelerating time-to-market and increasing energy efficiency of the massive data centers that run 24/7 at the semiconductor manufacturing process.
“The chip industry is the foundation of almost every other industry in the world,” says Huang. “With lithography at the frontiers of physics, NVIDIA’s introduction of cuLitho and collaboration with our partners TSMC, ASML and Synopsys enables fabs to increase throughput, reduce their environmental footprint and lay the groundwork for 2nm and beyond.”
Finally, the company also announced partnerships with Medtronic and Microsoft. The first, he said, will lead to the development of a common AI platform for software-defined medical devices that can improve patient care. Meanwhile, the latest Microsoft Azure host will see Nvidia Omniverse and Nvidia DGX Cloud.
The 2023 Nvidia GTC event runs through March 23.
VentureBeat’s mission is to become a digital city plaza where tech decision makers can learn about transformative business technology and execute transactions. Discover our Briefings.