Unveiled in April, H100 is constructed with 80 billion transistors and advantages from a variety of expertise breakthroughs. Among them are the highly effective new Transformer Engine and an NVIDIA NVLink® interconnect to speed up the biggest AI fashions, like superior recommender programs and huge language fashions, and to drive improvements in such fields as conversational AI and drug discovery.
“Hopper is the new engine of AI factories, 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,” mentioned Jensen Huang, founder and CEO of NVIDIA. “Hopper’s Transformer Engine boosts performance up to an order of magnitude, putting large-scale AI and HPC within reach of companies and researchers.”
In addition to Hopper’s structure and Transformer Engine, a number of different key improvements energy the H100 GPU to ship the following huge leap in NVIDIA’s accelerated compute information heart platform, together with second-generation Multi-Instance GPU, confidential computing, fourth-generation NVIDIA NVLink and DPX Instructions.
A five-year license for the NVIDIA AI Enterprise software program suite is now included with H100 for mainstream servers. This optimizes the event and deployment of AI workflows and ensures organizations have entry to the AI frameworks and instruments wanted to construct AI chatbots, advice engines, imaginative and prescient AI and extra.
Global Rollout of Hopper
H100 allows firms to slash prices for deploying AI, delivering the identical AI efficiency with 3.5x extra vitality effectivity and 3x decrease whole price of possession, whereas utilizing 5x fewer server nodes over the earlier era.
For clients who wish to instantly attempt the brand new expertise, NVIDIA introduced that H100 on Dell PowerEdge servers is now obtainable on NVIDIA LaunchPad, which gives free hands-on labs, giving firms entry to the newest {hardware} and NVIDIA AI software program.
Customers also can start ordering NVIDIA DGX™ H100 programs, which embody eight H100 GPUs and ship 32 petaflops of efficiency at FP8 precision. NVIDIA Base Command™ and NVIDIA AI Enterprise software program energy each DGX system, enabling deployments from a single node to an NVIDIA DGX SuperPOD™ supporting superior AI improvement of huge language fashions and different huge workloads.
H100-powered programs from the world’s main laptop makers are anticipated to ship in the approaching weeks, with over 50 server fashions in the market by the top of the yr and dozens extra in the primary half of 2023. Partners constructing programs embody Atos, Cisco, Dell Technologies, Fujitsu, GIGABYTE, Hewlett Packard Enterprise, Lenovo and Supermicro.
Additionally, a few of the world’s main greater training and analysis establishments shall be utilizing H100 to energy their next-generation supercomputers. Among them are the Barcelona Supercomputing Center, Los Alamos National Lab, Swiss National Supercomputing Centre (CSCS), Texas Advanced Computing Center and the University of Tsukuba.
H100 Coming to the Cloud
Amazon Web Services, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure shall be among the many first to deploy H100-based cases in the cloud beginning subsequent yr.
“We look forward to enabling the next generation of AI models on the latest H100 GPUs in Microsoft Azure,” mentioned Nidhi Chappell, common supervisor of Azure AI Infrastructure. “With the advancements in Hopper architecture coupled with our investments in Azure AI supercomputing, we’ll be able to help accelerate the development of AI worldwide.”
“By offering our customers the latest H100 GPUs from NVIDIA, we’re helping them accelerate their most complex machine learning and HPC workloads,” mentioned Karan Batta, vp of product administration at Oracle Cloud Infrastructure. “Additionally, using NVIDIA’s next generation of H100 GPUs allows us to support our demanding internal workloads and helps our mutual customers with breakthroughs across healthcare, autonomous vehicles, robotics and IoT.”
NVIDIA Software Support
The superior Transformer Engine expertise of H100 allows enterprises to shortly develop massive language fashions with a better stage of accuracy. As these fashions proceed to develop in scale, so does the complexity, typically requiring months to coach.
To sort out this, a few of the world’s main massive language mannequin and deep studying frameworks are being optimized on H100, together with NVIDIA NeMo Megatron, Microsoft DeepSpeed, Google JAX, PyTorch, TensorMovement and XLA. These frameworks mixed with Hopper structure will considerably pace up AI efficiency to assist prepare massive language fashions inside days or hours.
To study extra about NVIDIA Hopper and H100, watch Huang’s GTC keynote. Register for GTC free of charge to attend classes with NVIDIA and business leaders.
