Our mission to improve business is fueled by readers like you. To enjoy unlimited access to our journalism, subscribe today.
Star chip designer Jim Keller has joined Tenstorrent, a Toronto startup that manufactures specialized computer processors for artificial intelligence.
Keller’s career is filled with short but meaningful stays in large tech companies. Its imprint on Intel, which has faces recent product delays and has struggled to reclaim its industry dominance after missing the boat on mobile computing, may not be evident for a few more years.
Achieve a zen state
At Tenstorrent, Keller will take on the titles of President and CTO. The veteran chip genius will be tasked with guiding the nearly five-year-old company toward its goal of creating chips to train machine learning programs that go beyond current designs. Tenstorrent’s strategy is to use software to better distribute the use of computing power on each chip.
Ljubisa Bajic, co-founder and CEO of Tenstorrent, has spent more than a decade designing chips at AMD. His tenure overlapped with that of Keller when Keller helped design AMD’s now successful Zen Architecture from 2012 to 2015. Bajic also worked at Nvidia and VLSI.
Keller’s implicit endorsement could help the startup attract more investment and industrial partners. “Tenstorrent has made impressive progress, and with the most promising architecture on the market, we are poised to become a next generation computing giant,” Keller said in a statement.
The race to create better chips for AI applications is crowded. Competitors range from established chip giants AMD, Intel, and Nvidia to a host of startups, including GraphCore and Run.AI.
Every business has a different approach. Nvidia’s Ampere AI chip seeks to multitask with 54 billion transistors, the microscopic building blocks of semiconductor circuits. California start-up Cerebras makes physically huge chips, with 1.2 trillion transistors each. Putting more transistors on a chip can increase computing power, but also consume more power and complicate design in other ways.
Tenstorrent claims that its Greyskull processor introduced last year is capable of 368 trillion operations per second on AI tasks. That’s more than the 130-260 billion operations claimed by Nvida’s Tesla T4 system, although there are many variables that go into how a chip performs in real-world applications.
More to read absolutely technological coverage of Fortune:
- A brief history of Bitcoin bubbles
- Commentary: Why the United States Needs a national climate investment fund
- the the biggest conspiracy theories of 2020 (and why they won’t die)
- Why Julian Assange’s victory does little to help the cause of press freedom
- Comment: Marketers Can’t Predict What You’ll Buy– Even if they use AI