Tech
Computer Graphics
Ramblings on Nvidia's acquisition of ARM
Feb 10, 2021

Nvidia has recently acquired a majority stake in ARM, in a $40 billion dollar deal from Softbank that’s set to make waves in the world of computing for decades to come and easily even further.
Why does it matter, you ask? First, a short history of both ARM and Nvidia for context.
ARM Limited is a British semiconductor design company (originally Acorn RISC Machine) with a lion’s share in mobile and embedded processing. I mention this because their designs sit at the core of almost every smartphone on the planet. Apple, Samsung, Huawei, Qualcomm and Mediatek all rely on ARM designs to manufacture their processors. All that without considering the embedded devices like the Raspberry Pi and server processors like Ampere, the first 80 core ARM based processor being used in the fastest supercomputer known to man to date. Intellectual property that now falls onto Nvidia to decide the fate of.
Now onto the company that is looking to pioneer artificial intelligence and push it into the future, Nvidia. Originally they made their name designing gaming graphics cards but caught their billion, arguably to come a trillion dollar break with the ‘coincidence’ of how well their gaming GPUs handled parallel computing. This allowed them to move into the data center and server market, and recently into the mobile and embedded market with their Tegra mobile processors and the Jetson nano.
With the history lesson over, let’s talk about company culture and philosophy. ARM has an open model, licensing their designs to anyone interested. Nvidia on the other hand are famous for their lack of hardware documentation and closed nature of their entire product stack. From that alone, there is a conflict.
Apple silicon, the heart of Apple’s business is based on ARM and to put it lightly, their relationship with Nvidia is dicey. Broadcom, the manufacturers of the system on a chip(SOC) that sits at the heart of the Raspberry Pi also have a tenuous relationship with Nvidia. Linus Trovalds, the brain behind Linux has publicly flipped off Nvidia at an interview for how difficult they are to work with.
Corporate conflict aside, the most pertinent topic is the one in your hands right now, your smartphone. Devices so integral to modern day life they drive all aspects of your life. All modern smartphones have ARM designs running at their core and how will the precedent Nvidia has set affect ARM’s business model moving forward?
While I’d love to answer the question and speculate on the future of ARM, the meat and potatoes are the other players in the processor market.
At the risk of oversimplifying, a processor is essentially a fancy calculator. 2*4 is the same as 2+2+2+2 and therein lies the difference between Reduced instruction Set Computers(RISC) and Complex Instruction Set Computers(CISC). One has more instructions to crunch numbers with and the other not so much. You can perform the same operations just that a RISC computer would need more operations to the same thing as a CISC computer
Intel has been the pioneer of CISC with its x86 instruction set that runs most desktop and laptop computers today. However, with compute shifting steadily to the cloud, mobile processors require less and less horsepower(CISC’s strength) and require more power efficiency(RISC’s strength). My point? We just might be looking at a shift in paradigm with how computing is handled. Where most users are constantly connected and compute is handled in the cloud and only niche users require compute intensive workstations. The shift is already happening with phones being the primary compute devices and some users having laptops and even fewer desktop PCs.
Nvidia also recently completed its acquisition of Melanox, an Israeli networking company and has used that technology to build massive compute servers with its networked Ampere architecture compute cards. A sign of their future market plan and the times to come.
Lastly RISC-V, an entirely open-sourced CPU instruction set architecture that has been making huge leaps and strides. While not close enough to be a competitor to ARM, x86 or PowerPC it holds great promise for a free, accessible and hopefully competitive instruction set architecture. Oh, and also Power PC, IBM’s brainchild has recently been open sourced adding onto the kerfuffle and promising more competition in the space.
UPDATE!! They did not acquire ARM due to anti trust concerns being raised. Which, for me, is a relief.