Nvidia announced today that the Nvidia H100 Tensor Core graphics processing unit (GPU) is in full production, and that international technology partners would launch the first wave of products and services based on the Nvidia Hopper architecture in October.
Jensen Huang, Nvidia's CEO, made the announcement during the Nvidia online GTC fall event.
H100, which was unveiled in April, is built with 80 billion transistors and contains a slew of technological advancements, including a powerful new Transformer Engine and an Nvidia NVLink interconnect that will help accelerate the largest artificial intelligence (AI) models, such as advanced recommender systems and large language models, as well as driving innovations in areas such as conversational AI and drug discovery.
"Hopper is the primary AI engine, processing and refining mountains of data to train models with trillions of parameters that are used to drive advances in language-based AI, robotics, healthcare, and life sciences," according to Jensen Huang, Nvidia's founder and CEO.
The H100 GPU is powered by several other powerful innovations, including a second-generation Multi-Instance GPU, confidential computing, and fourth-generation Nvidia NVLink and DPX instructions.
In a press conference, Nvidia's general manager of accelerated computing said the H100 is now in full production. "We're ready to take orders for shipment in Q1, from Nvidia's fiscal year in October." And starting next month, our systems partners from Asus to Supermicro will begin shipping their H100 systems, beginning with the PCIe products and expanding later this year to the NVLink HDX platforms."
H100 for mainstream servers now includes a five-year license for Nvidia's AI Enterprise software suite. This simplifies the development and deployment of AI workflows and guarantees that businesses have access to the necessary AI frameworks and tools to build AI chatbots, recommendation engines, vision AI, and more.
Hopper's global adoption is expected to begin in the coming weeks.
H100 helps businesses lower expenses for using artificial intelligence, while achieving the same level of performance as the previous generation, but with a five-fold reduction in server nodes.
Nvidia has announced that H100 on Dell PowerEdge servers will now be available on Nvidia LaunchPad, a free hands-on lab that allows corporations to try out the latest Nvidia AI software.
Customers may also order Nvidia DGX H100 systems, which include eight H100 GPUs and deliver 32 petaflops of performance at FP8 precision. Every DGX system is powered by Nvidia Base Command and Nvidia AI Enterprise software, which enable deployments from a single node to an Nvidia DGX SuperPOD, supporting advanced AI development of large language models and other massive workloads.
H100-powered systems from the world's leading computer manufacturers are expected to ship in the coming weeks, with over 50 server models in the market by the end of the year and dozens more in the first half of 2023. Partnering systems include Atos, Cisco, Dell Technologies, Fujitsu, GIGABYTE, Hewlett Packard Enterprise, Lenovo, and Supermicro.
H100 will be used by some of the world's leading higher education and research organizations to power their next-generation supercomputers. Among them are the Barcelona Supercomputing Center, Los Alamos National Lab, the Swiss National Supercomputing Centre (CSCS), the Texas Advanced Computing Center, and the University of Tsukuba.
Buck noted that the previous A100 generation had 320 A100 systems in a datacenter, but with Hopper a data center would only need 64 H100 systems to match that throughput of the previous data center. That's a 20% decrease in nodes and a huge improvement in energy efficiency.