Nvidia Announces New CPU R&D Effort in Israel
Nvidia has been on quite a spending spree as of late, thanks to the pandemic-related shortages that have resulted in record breaking revenue for the famed GPU company. Not only is it trying to drop a cool $40 billion to scoop up Arm Ltd., but a new report from an Israeli publication named Globes indicates the company is expanding its CPU-related endeavors in Israel, which is where Intel has had a home base for years. You don’t need to be a mathlete to figure this one out, as the company is looking to hire CPU engineers, and is doing so where Intel currently operates. The beeping sound you hear in the distance is the Nvidia cash truck backing up to the loading dock at Intel’s Israeli HQ, in other words.
The news report states that Nvidia is looking to beef up its CPU team in Israel, which already stands at 2,800 engineers strong, and it will do so by creating a new design and engineering group to help craft next-generation Nvidia CPUs. In a statement, Nvidia CTO Michael Kagan said, “Israel, with its unique wealth of talent, is a key player in the global tech ecosystem, and we are excited to be creating a new CPU group here.” He further explained the company’s effort would be centered around, “growing our local R&D activities both in this area and in our extensive work supporting the local ecosystem through unique programs for startups and developers.” Interestingly, Mr. Kagan was at Intel for 16 years before co-founding Israeli interconnect company Mellanox, which was purchased by Nvidia in 2020 for $6.9 billion.
The move by Nvidia makes it clear that it’s serious about ramping up its CPU business, which is semi-new territory for the GPU company. While Nvidia has built licensed ARM cores and its own custom microarchitecture for over a decade, it has never had a presence in the kind of high-end markets where the company is now thought to be focusing its efforts. Nvidia’s plans are to take on the data center CPU business, as opposed to building chips for gamers and home users who currently use Intel and AMD processors. In fact, Nvidia itself currently uses AMD Epyc CPUs in some of its data center products, with its DGX STATION A100 using an AMD 7742 EPYC chip to link four Nvidia A100 GPUs.
Nvidia has been working on its own CPU for awhile now, and announced its first Arm-based chip code-named Grace in 2021. Grace is a CPU designed for AI and HPC workloads, and Nvidia states it was created because these workloads demand, “a tight coupling of the CPU and GPU.” The company noted in its release that with this “tight coupling,” a Grace CPU-based system could be up to 10x faster than its own DGX-based systems, which as you recall run on x86 AMD CPUs.
News of Nvidia ramping up its CPU efforts comes as its acquisition of Arm is hanging in the balance, with the latest development being a lawsuit filed by the Federal Trade Commission (FTC) to stop what it calls “the largest semiconductor chip merger in history.” The FTC is arguing that buying Arm would “give the company both “means and incentive” to stifle the development of various technologies.” Nvidia is also facing difficulties in the UK, where its chief of Digital Security has announced an intervention based on concerns about national security. To put it bluntly, it’s not looking good for Nvidia’s Arm acquisition at this point, but luckily the company has record breaking revenue as a consolation prize.