This article is compiled by Semiconductor Industry (ID: ICVIEWS)

Tenstorrent's second-generation all-purpose AI processor should be available by 2025.

Nvidia dominates the artificial intelligence (AI) processor market, controlling more than 80% of sales, according to some recent estimates. But Jim Keller, a legendary processor designer and current CEO of Tenstorrent, believes that some markets are not well served by Nvidia. As a result, there is an opportunity for Tenstorrent and other AI processor developers.

"Nvidia doesn't serve many markets well," Jim Keller said in an interview with Japanese media.

Jim Keller has a truly distinguished history in the computer industry, making waves at AMD, Intel, and Tesla, and is now leading the development of AI processors at Tenstorrent. From 1998 to 1999, Jim Keller worked on the K7/K8 architecture that supported Athlon at AMD, from 2008 to 2012, he led the development of the A4 and A5 processors at Apple, from 2012 to 2015, he presided over the K12 ARM project and the Zen architecture project at AMD, from 2016 to 2018, he developed the FSD autonomous driving chip at Tesla, and from 2018 to 2020, he participated in a mysterious project at Intel.

Jim Keller's leadership at Tenstorrent aims to provide an affordable alternative to Nvidia's expensive GPUs, which cost $20,000 to $30,000 or more each. Tenstorrent's business approach is designed to serve markets that Nvidia has not adequately addressed, especially at the edge. Tenstorrent claims that its Galaxy system is three times more efficient and 33% less expensive than Nvidia's DGX, which is probably the most popular AI server in the world.

Tenstorrent is reportedly on track to release its second-generation multi-purpose AI processor by the end of this year, but did not reveal the name of the processor. According to Tenstorrent’s latest roadmap last fall, the company intends to release its Black Hole standalone AI processor and Quasar low-power, low-cost chiplets for multi-chiplet AI solutions.

The company claims that its upcoming processors offer performance efficiency comparable to Nvidia's AI GPUs. The touted efficiency and lower cost are achieved in part by avoiding high-bandwidth memory (HBM) in favor of GDDR6, which is a logical thing to do for entry-level and mainstream AI processors designed primarily for AI inference. At the same time, Tenstorrent says its architecture consumes less memory bandwidth than its competitors, which is a key reason for its higher efficiency and lower cost.

While Tenstorrent has yet to capture a significant share of the AI ​​processor market, the company is currently focused on cost-effective and scalable AI solutions that can address a variety of applications that Nvidia currently cannot properly serve. However, these markets will not be blue oceans, as many companies will try to address these markets with their products in the coming quarters, fighting against the entrenched Nvidia. Rather than competing head-on with Nvidia, it will be easier for new entrants to address niche markets that are indeed not directly catered to by the "green team."

In early 2023, Canadian AI chip startup Tenstorrent announced personnel adjustments, and CTO Jim Keller will take over as CEO.

Interestingly, former CEO Ljubisa Bajic swapped jobs with Keller and will serve as the new CTO.

Keller wrote on the social platform, "Ljubisa Bajic and I have decided to swap roles and continue to develop artificial intelligence, CPU products, etc. to meet the next generation of computing challenges."

Tenstorrent was founded seven years ago, and Bajic is one of the co-founders. Jim Keller met Bajic while working at AMD and became the first external investor in Tenstorrent. He then officially joined Tenstorrent as CTO in January 2021.

Keller had two main tasks during his tenure as CTO. One was to bring in the RISC-V architecture, based on which the company team developed its own Ascalon CPU core. The other was to bring in David Bennett as Chief Customer Officer, an AMD veteran who had also served as Lenovo Japan President and NEC PC Business CEO.

However, Ascalon's commercial products will still have to wait for some time. Tenstorrent's new generation BlackHole AI chip will be developed based on SiFive's X280 RISC-V core design.

Jim Keller suggests Nvidia abandon private standards

Jim Keller is not only a chip design master with great technical skills, but also a loyal supporter of open technology. He has always hated closed technologies. Naturally, Nvidia became his "enemy".

Recently, Jim Keller proposed that Nvidia's latest Blackwell GPU should not use the private NVLink standard protocol for multi-chip interconnection and network interconnection, but should switch to the open Ethernet standard, which can save Nvidia billions of dollars. He also believes that Nvidia should not use its own solution InfiniBand in data center networks, but should also switch to Ethernet.

Although NVIDIA Infiniband networks have low latency and high bandwidth, up to 200GbE, Ethernet can achieve 400GbE or even 800GbE.

Giants such as AMD, Broadcom, Intel, Meta, Microsoft, and Oracle are also working together to develop the next-generation ultra-high-speed Ethernet (Utlra Ethernet), which has higher throughput and is more suitable for AI and HPC applications.

In addition, Jim Keller has always been dissatisfied with NVIDIA CUDA's closed ecosystem, and once called it a swamp rather than a moat.

Arm, Intel, Qualcomm, Samsung and others have also cooperated to form the Unified Accelerator Foundation (UXL), one of the goals of which is to replace NVIDIA's solution. For Huang Renxun, both NVLink and CUDA are the results of his company's investment of billions of dollars over many years, and are also the tools to protect his own interests.

*Disclaimer: This article is written by the original author. The content of the article is his personal opinion. We reprint it only for sharing and discussion, and it does not mean that we agree or agree. If you have any objection, please contact the backstage.