NextSilicon reveals new processor chip in challenge to Intel, AMD

Semiconductor chips are seen on a printed circuit board in this illustration picture taken February 17, 2023. REUTERS/Florence Lo/Illustration · Reuters

By Stephen Nellis

SAN FRANCISCO (Reuters) -NextSilicon, an Israeli startup whose computing chips are being evaluated by U.S. national labs, on Wednesday said it is developing a central processor that it hopes will rival Intel and AMD and help it compete with Nvidia‘s systems.

NextSilicon’s has raised $300 million in funding and its flagship “Maverick-2” chip is designed to speed up precision scientific computing tasks such as modeling nuclear weapons. That field was once dominated by Nvidia, but as Nvidia has focused its attention on lower-precision computing tasks such as artificial intelligence, startups like NextSilicon have tried to take advantage of the AI giant’s shift.

On Wednesday, NextSilicon also disclosed for the first time that it is developing a complement to its main chip in the form of a new central processing unit, a market still dominated by Intel and AMD. It uses technology called RISC-V, an open computing standard that competes with Arm Ltd and is increasingly being used by chip giants such as Nvidia and Broadcom.

Nvidia often pairs its chip with either its own or a third-party central processing unit, even partnering with companies such as Intel to create a tighter coupling between the two classes of chips.

At present, NextSilicon said its central processing unit remains a test chip. But its Maverick-2 chips are in production, and NextSilicon claims it can carry out some of the same work as Nvidia chips, but faster and using less power and without rewriting the software code used. Sandia National Laboratories has been evaluating prototype systems made with NextSilicon’s chips for three years.

James H. Laros III, senior scientist and Vanguard program lead at Sandia National Laboratories, said in a statement that NextSilicon’s “performance results are impressive, showing real promise for advancing our computational capabilities without the overhead of extensive code modifications.”

(Reporting by Stephen Nellis in San Francisco; Editing by Stephen Coates)