Arm has been central to a series of announcements over the past month — including a blockbuster acquisition offer by NVIDIA that I detailed here — all of which confirm what many have long seen coming, that Intel has far more to worry about than just AMD. From its origins as an experimental RISC design, Arm grew into the de facto standard for mobile device SoCs.
Over the past decade, as the release cadence and performance slope of new Intel CPUs has slowed, Arm-based designs have closed the speed and functionality gap as licensees like Apple and Qualcomm have packed the latest mobile phones with technology that leaves two-year-old PCs in the dust. The path of Arm from the lilliputian designer of low-end embedded components to a disruptive challenger to Intel's processor hegemony was marked by two recent milestones:
- AWS acquired Annapurna Labs and using its Arm expertise to develop a custom Arm CPU suitable for SmartNICs and cloud computing workloads. AWS rapidly evolved the technology into a second-generation Graviton 2 processor that is superior to x86 alternatives in many situations.
- Apple, like AWS-Annapurna, a licensee of the Arm architecture and developer of the most powerful mobile SoCs available, announced its switch to Arm-based Apple Silicon with a family of SoCs for all future Mac models to "create a common architecture across all Apple products."
AWS's endorsement and successful implementation of Arm as an option for EC2 compute instances finally broke the curse of Arm in the data center, an idea that had been unsuccessfully tried several times over the past few years. Arm then cemented itself as a viable option in the data center with the Neoverse N1 announcement in 2019, promising "the first Arm platform specifically designed from the ground up for infrastructure, on a roadmap committed to delivering more than 30% higher performance per generation."
Arm's umbrella stretches from cloud to edge
As Apple demonstrated, the Arm architecture is both extremely efficient and highly scalable, making it suitable for a range of applications that stretches from smartwatches to supercomputers. At its recent Developer Summit, Arm updated a roadmap showing a performance curve that makes Arm a viable, and in some cases, preferred, alternative to x86 processors, yet scan still scales down to compact, power-efficient edge implementations.
Arm previewed a cloud-to-edge roadmap that bifurcates its next-generation Neoverse products into two designs:
- The V1 optimized for single-threaded 64-bit performance more than 50 percent higher than the N1, but adding Scalable Vector Extensions (SVE) to accelerate machine/deep learning, HPC and other applications using matrix math operations.
- The N2 for a variety of cloud and carrier scenarios including scale-out, mobile edge and SmartNic applications and targeting a 30-40% performance improvement over the N1.
Arm reiterated its three-pronged approach to facilitate developers and system designers. As the company's client LoB VP writes (emphasis added):
We built Total Compute around three key principles: increasing compute performance of the system, better developer access to that performance through our software and tools, and security protection throughout the ecosystem. Each of these pillars is essential to our vision for the future of compute, one that expands the developer’s world with new capabilities to unlock innovation and enable the highest levels of creativity.
Recent enhancements include:
- Chip-level interfaces, including CCIX and CXL, to enable design partners to add hardware features and connect with networked devices, along with platform security modules to prevent attacks on system firmware and OSs.
- Developer tools including added support for industry standards, popular operating systems, developer environments and reference designs. Improved viability as a cloud platform by working with partners on Arm-compatible implementations of popular hypervisors like Xen and KVM and containers software including Docker and Kubernetes.
- A commitment to an exclusively 64-bit architecture for all future "big" cores, i.e. the high-performance half of its big.LITTLE system architecture, by 2022.
Arm disrupted client computing by starting small (phones) and expanding to more demanding and general-purpose applications in tablets and PCs. It takes the same strategy of efficiency, scalability, flexibility and customization to the attacking the needs of enterprises, cloud operators and carriers with a consistent processor architecture that can be adapted to everything between far edge (think IoT devices and SmartNICs) and shared cloud infrastructure.
The updated Neoverse products, along with the AWS Graviton 2, demonstrate Arm's ability to deliver comparable performance to x86 systems in many situations with superior density and power efficiency. However, as mobile SoCs have long demonstrated, Arm's extensive support for system integrations, makes it a better approach for bespoke SoCs tailored to particular workloads including AI, network and security services, IoT data processing and HPC.
Chris Bergey, Arm's SVP of infrastructure products frames Arm as a disruptive technology based on choice versus standardization (emphasis added):
In infrastructure, we’re seeing a quiet revolution in how we deliver and distribute compute. It can be distilled down to a single word: choice. The days of homogeneous one-size-fits-all, server farms powered by a single, legacy, general-purpose compute architecture are being displaced by solutions that allow greater vendor choice and flexibility in how to distribute processing at optimal points along the compute spectrum from cloud to edge. In this way, the right resources can be layered in at the right points along the spectrum.
Intel has seen the revolution coming and now does a robust business customizing its standard SKUs for various cloud builders. Indeed, two years ago the company's cloud service provider business said that half of its data center processors, i.e. Xeon, were tweaked to meet the requirements of its largest cloud customers (emphasis added):
50% of our [processor] volume is optimized spanning over 10 different cloud service providers and nearly 30 different SKUs in addition to the SKUs that you saw that cover our standard road map. These are off-road map SKUs because they're customized uniquely for the service provider and they provide differentiation so it's unique to them and they want to keep it off road map.
Nonetheless, Arm has the advantage with an extensive library of standard modules and cores, partner ecosystem and more flexible IP licensing model (which, in the hands of someone like Apple can be used for further product differentiation).
While nature abhors a vacuum, the technology world abhors a monopoly and after decades of dominance, Intel's monopoly faces a two-front attack, from AMD and Arm-NVIDIA. There will be many exciting hardware battles to watch over the coming years, but unlike most wars, cloud and enterprise customers won't be the collateral damage, but rather the beneficiaries.