April 8, 2024

Nvidia unveils latest chips, technology to speed up AI computing

Nvidia unveils latest chips, technology to speed up AI computing

The brand of technologies corporation Nvidia is observed at its headquarters in Santa Clara, California, February 11, 2015. REUTERS/Robert Galbraith

Sign up now for Totally free limitless access to Reuters.com

March 22 (Reuters) – Nvidia Corp (NVDA.O) on Tuesday declared quite a few new chips and systems that it explained will enhance the computing velocity of progressively intricate artificial intelligence algorithms, stepping up competitors from rival chipmakers vying for valuable info center organization.

Nvidia’s graphic chips (GPU), which originally served propel and enrich the quality of films in the gaming current market, have turn out to be the dominant chips for providers to use for AI workloads. The most up-to-date GPU, termed the H100, can enable decrease computing situations from weeks to times for some function involving schooling AI styles, the business reported.

The announcements have been built at Nvidia’s AI builders convention online.

Sign up now for Free endless obtain to Reuters.com

“Info centers are becoming AI factories — processing and refining mountains of knowledge to make intelligence,” claimed Nvidia Main Govt Officer Jensen Huang in a statement, calling the H100 chip the “engine” of AI infrastructure.

Firms have been employing AI and equipment learning for every little thing from making recommendations of the upcoming online video to check out to new drug discovery, and the technological innovation is progressively turning into an critical software for organization.

The H100 chip will be generated on Taiwan Production Semiconductor Firm’s (2330.TW) reducing edge four nanometer procedure with 80 billion transistors and will be available in the 3rd quarter, Nvidia claimed.

The H100 will also be used to establish Nvidia’s new “Eos” supercomputer, which Nvidia reported will be the world’s swiftest AI system when it begins operation later this year.

Facebook parent Meta (FB.O) declared in January that it would make the world’s swiftest AI supercomputer this calendar year and it would complete at practically 5 exaflops. Nvidia on Tuesday explained its supercomputer will run at in excess of 18 exaflops.

Exaflop overall performance is the capacity to carry out 1 quintillion – or 1,000,000,000,000,000,000 – calculations for each next.

In addition to the GPU chip, Nvidia launched a new processor chip (CPU) called the Grace CPU Superchip that is primarily based on Arm technology. It is the first new chip by Nvidia based on the Arm architecture to be declared because the firm’s offer to purchase Arm Ltd fell aside final month due to regulatory hurdles.

The Grace CPU Superchip, which will be out there in the initial 50 {18fa003f91e59da06650ea58ab756635467abbb80a253ef708fe12b10efb8add} of following 12 months, connects two CPU chips and will aim on AI and other jobs that demand intense computing energy.

Much more corporations are connecting chips using technological innovation that allows a lot quicker facts movement involving them. Before this month Apple Inc unveiled its M1 Ultra chip connecting two M1 Max chips. [nL2N2VB1DI]

Nvidia stated the two CPU chips ended up linked working with its NVLink-C2C technological innovation, which was also unveiled on Tuesday.

Nvidia shares were being up extra than 1{18fa003f91e59da06650ea58ab756635467abbb80a253ef708fe12b10efb8add} in midday trade.

Sign up now for Totally free unrestricted obtain to Reuters.com

Reporting By Jane Lanhee Lee, more reporting by Joseph White Editing by Bernard Orr

Our Specifications: The Thomson Reuters Trust Principles.