Nvidia is flexing its arm – the company has entered into the CPU business. Building new Arm-based CPUs, they promise ten times the performance of today’s fastest servers for AI and HPC workloads.

Arming research with enhanced supercomputing powers.

Two HPC centers at CSCS and Los Alamos will house the first two Nvidia Grace-based supercomputers (named after computing pioneer Grace Hopper), with Hewlett-Packard Enterprise (HPE) as the system provider. Target deliveries are set for 2023.

The intended CSCS “Alps” HPC system is slated to be the world’s most powerful AI-capable supercomputer. Delivering a whopping 20 exaflops of AI performance – equivalent to one quintillion FLOPS – it’ll be based on Nvidia’s mixed-precision arithmetic and sparsity features. Researchers in the Switzerland institute await with high anticipation as the new system will undoubtedly enable breakthroughs in a plethora of scientific research fields from quantum chemistry to astrophysics.

The diverse mission needs of the laboratory at Los Alamos will also be met once it receives the Nvidia Grace Central Processing Unit (CPU). “Advanced computational tools provide a powerful enabler to help advance scientific discovery and use across all our applications, ranging from climate and disease modeling to materials science, to nuclear deterrence and monitoring”, said John Sarrao, deputy director for Science, Technology and Engineering at Los Alamos.

Arm-based CPUs will soon be taking over the computing world, but let’s take a trip down memory lane to appreciate the history of this revolutionary invention.

An invention that shaped the modern world.

For you Arm geeks, Steve Furber and Sophie Wilson are the names that should be engraved in your minds. These two very talented computer scientists were the principal designers of the Arm architecture. Given the task of fabricating the microprocessor design for Acorn’s own 32-bit processor with little to no resources, Steve worked on the chip design while Sophie developed the instruction set for the Arm1, which eventually led to the birth of the Acorn Archimedes, the first RISC-based home computer.

Originally an acronym for “Acorn RISC Machines”, what ‘Arm’ stood for changed to “Advanced RISC Machines” after the company Arm Ltd. was founded in 1990. Arm Ltd is a joint venture between Acorn Computers, Apple Computer (now Apple Inc.) and VLSI Technology. RISC is an acronym for reduced instruction set computing, which is vastly different from Intel’s x86 architecture. The IT sector sure does love acronyms of acronyms!

A key difference between Arm and other chip companies such as Intel or TSMC, is that Arm doesn’t actually manufacture anything. Arm is what we call a fabless chip design company. It develops the architecture of processors and licenses this intellectual property to other companies such as Samsung or Qualcomm. These companies in turn design their own products and incorporate the Arm architecture into their own chips, and then sell the components to Apple, Tesla or other final end users.

Compare this to an author making money from a book publisher. You get paid by Scholastic based on how many books you sell through online platforms, but apart from the manuscript, you have zero control over anything else. In the same scenario, Arm receives both licence fees and royalties based on eventual production volumes.

Arm is everywhere.

Arm licenses IP, CPU, GPU designs to many companies, including almost every smartphone chip maker thanks to its very low power consumption – a particularly important aspect for today’s smartphones. That includes Apple, which has Arm-compatible CPUs in almost all its products, such as the A14 chip for the latest iPhones and the M1 chip for Mac computers.

A non-exhaustive list of other Arm-incorporated products include Fujitsu’s A64FX microprocessor (powers the fastest supercomputer in the world, Fugaku), Qualcomm’s Snapdragon CPU (Android devices and Windows laptops), Samsung’s Exynos system-on-chips (SoC), Broadcom’s Raspberry Pi, Cavium’s ThunderX Arm server microprocessors, Nvidia’s Tegra SoC series for smartphones, and the Chinese tech giant, Huawei’s HiSilicon smartphone processors. Even the Intelligent Platform Management Interface (IPMI) architecture of your computer motherboard has Arm technology built in its baseboard management controller (BMC).

Arm-based products are evidently ubiquitous. Any ‘smart’ device is likely to contain Arm technology inside. Being the most popular microprocessor architecture, Arm reported its silicon partners shipped a record 6.7 billion Arm-based chips in the fourth quarter of 2020, more than other popular CPU instruction set architectures combined. That’s 842 chips shipped per second!

While Arm has been the preferred CPU for virtually all smartphones, a Nvidia-Arm acquisition will have astounding impacts on the broader computing industry. Expanding into cloud data centres and HPC installations, Arm recently announced new details for their Neoverse platform, placing targets on growing markets such as hyperscale cloud operators, HPC clusters, 5G infrastructure and IoT systems. 

Still not a done deal and with many hurdles to clear, this acquisition could help expose Arm’s enormous customer base to the power of GPU computing, while allowing Nvidia to build better integrated “end-to-end” systems. With our increasing reliance on accelerated computing to resolve some of the globe’s biggest problems such as designing self-driving vehicles, curing COVID-19, or exploring space in search for the origins of life, a technological marriage between Nvidia-Arm will surely put some of humankind’s complex quests to an end.