Shulou(Shulou.com)11/24 Report--

CTOnews.com, November 13 (Xinhua)-- Nvidia today released the next generation of artificial intelligence supercomputer chips that will play an important role in deep learning and large language models (LLM), such as OpenAI's GPT-4. The new chip represents a significant leap over the previous generation and will be used in data centers and supercomputers to handle tasks such as weather and climate prediction, drug discovery and quantum computing.

The key product of this release is the HGX H200 GPU based on Nvidia's "Hopper" architecture, which is the successor to the H100 GPU and the company's first chip to use HBM3e memory, which is faster and larger, so it is more suitable for large language models. The performance has been directly improved by 60% to 90% compared with the performance of the former overlord H100Mague H200. "with HBM3e, the Nvidia H200 provides 141GB memory at 4.8 TB per second, almost twice the capacity and 2.4 times the bandwidth compared to the A100," Nvidia said.

In terms of artificial intelligence, Nvidia says the HGX H200 reasoning on Llama 2 (70 billion parameter LLM) is twice as fast as H100. The HGX H200 will be available in 4-way and 8-way configurations and is compatible with software and hardware in H100 systems. It will be available for every type of data center (on premises, cloud, hybrid cloud, and edge) and will be deployed by Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure, and will be launched in the second quarter of 2024.

Another key product of Nvidia's launch is the GH200 Grace Hopper Super Chip (superchip), which combines the HGX H200 GPU and the Arm-based Nvidia Grace CPU through the company's NVLink-C2C interconnect, which officials say is designed for supercomputers, allowing scientists and researchers to solve the world's most challenging problems by accelerating the run of complex AI and HPC applications for TB-level data.

GH200 will be used for "more than 40 AI supercomputers from global research centers, system manufacturers and cloud providers", including Dell, Eviden, HPE, Lenovo, QCT and Supermicro. It is worth noting that HPE's Cray EX2500 supercomputer will use four-way GH200, which can be expanded to tens of thousands of Grace Hopper superchip nodes.

Perhaps the largest Grace Hopper supercomputer is the JUPITER at the J ü lich factory in Germany, which will become "the most powerful AI system in the world" when installed in 2024. It uses a liquid-cooled architecture and its enhancement module consists of nearly 24000 Nvidia GH200 super chips, which are interconnected through the Nvidia Quantum-2 InfiniBand network platform.

Nvidia said JUPITER would help make scientific breakthroughs in a number of areas, including climate and weather forecasts, generate high-resolution climate and weather simulations, and make interactive visualization. It will also be used for drug discovery, quantum computing and industrial engineering, many of which use customized Nvidia software solutions that simplify development but also make supercomputing teams dependent on Nvidia hardware.

CTOnews.com noted that last quarter, Nvidia achieved a record $10.32 billion in revenue (total revenue of $13.51 billion) in AI and data centers alone, an increase of 171% from a year ago, and Nvidia no doubt hopes that the new GPU and super chips will help it continue this trend.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.