Nvidia has unveiled the H200 chips, which is a big enchancment from H100, and is predicted to be out there from Q2 2024.
Software program and fabless large Nvidia Company (NASDAQ: NVDA) has unveiled a brand new graphics processing unit known as the H200. A brand new GPU improve from the H100, the Nvidia H200 is designed to cater to synthetic intelligence (AI) fashions which are at present the crux of the continuing push for AI.
In response to an official announcementthe H200 can deal with giant quantities of information for generative AI and course of heavy computing workloads. Nvidia says that the H200 is the primary GPU with HBM3e, offering a quicker and bigger reminiscence for dealing with generative AI and enormous language fashions (LLMs). The brand new GPU can ship 141 GB of reminiscence at 4.8 terabytes per second. On the necessity for brand spanking new and improved chips, Nvidia’s vice chairman of hyperscale and HPC, Ian Buck, mentioned:
“To create intelligence with generative AI and HPC functions, huge quantities of information have to be effectively processed at excessive pace utilizing giant, quick GPU reminiscence. With NVIDIA H200, the business’s main end-to-end AI supercomputing platform simply acquired quicker to resolve among the world’s most vital challenges.”
Coaching AI methods and fashions requires 1000’s of those chips working collectively. OpenAI used the H100 to coach its GPT-4, its most superior LLM.
Nvidia to Launch H200 Subsequent 12 months and Proceed H100 Manufacturing
There may be at present no indication of how a lot the H200 will value. Nevertheless, it’s anticipated to be nicely greater than the typical $30,000 worth level for the H100. In response to Nvidia spokesperson Kristin Uchiyama, the corporate’s companions set the pricing.
Nvidia expects that the H200 might be about twice as quick because the H100. In response to the announcement, the brand new GPU might be out there from cloud service suppliers and world system producers within the second quarter of 2024. Nvidia named Google Cloud, Amazon Net Providers, Oracle Cloud, and Microsoft Azure, as among the first cloud service suppliers that can deploy the chips from subsequent 12 months. Different named cloud suppliers are Vultr, Lambda, and CoreWeave.
Moreover, the H200 GPU might be appropriate with {hardware} and software program at present working H100 methods. Which means organizations can deploy the H200 in any kind of information middle, with out having to eliminate their present H100-compatible server methods.
Uchiyama has famous that Nvidia doesn’t intend to pause H100 manufacturing. In response to the corporate spokesperson, Nvidia will proceed so as to add total provide all through 2024.
Final month, Coinspeaker reported that Nvidia quietly started designing central processing items (CPUs) for Home windows PCs. Nvidia will reportedly use know-how from Arm Holdings Ltd, the chip designer owned by Softbank Group. The chips are set for launch in 2025.
Nvidia Partnerships
Nvidia can also be engaged on a number of partnerships, and is collaborating with Google Cloud to deploy new AI software program and infrastructure. As well as, the large chip designer partnered with electronics manufacturing large Foxconn to construct AI-run information facilities for various functions, together with use circumstances for electrical autos. Nvidia says that these AI factories will assist digitize the manufacturing and inspection processes for robotics and electrical autos. The plan can also be for Foxconn shoppers to make use of the AI service to coach industrial robots and autonomous machines.
subsequent
