Arm Presents Its Most Expansive Array of Processors for the Automotive Industry
Arm has unveiled a cutting-edge ensemble of Automotive Enhanced (AE) processors, engineered to propel the computational capabilities of the automotive realm. This new collection is specifically aimed at uses that harness artificial intelligence and those that are vital for vehicular safety. With the innovative Armv9 architecture at their core, these processors deliver performance on par with servers and boast improved AI functionalities. These processors have been developed by Arm to address the escalating requirements of contemporary software-defined vehicles.
Arm's new AE processors include a server-class, automotive-focused Neoverse CPU, two Armv9 A-class CPUs, an R-class CPU, and an image signal processor (ISP).
All About Circuits sat down with Dipti Vachani, Arm's SVP and GM of automotive, to learn more about the new set of AE processors and what they bring to the table.
Introducing the Enhanced Automotive (AE) Processor Portfolio
Central to the AE series is the Neoverse V3AE processor, heralding the inaugural introduction of Neoverse technology into the automotive domain. The V3AE is crafted for unparalleled server-level computational power, boasting a substantial improvement—over 50% in per-core performance—compared to its forerunner, the Cortex-A78AE. Embracing the state-of-the-art Armv9 architecture, the Neoverse V3AE comes decked with innovative security enhancements including memory tagging extensions, pointer authentication, branch target identification, along with support for robust hardware-based virtualization.
Vachani remarks on the evolution within the automotive sector, "In our interaction with industry partners, we recognize the escalating requirement for high-performance solutions driven by the surge in AI applications. The obliging and complex computations in autonomous scenarios necessitate substantial raw CPU muscle tailored for AI tasks. In response, we've channeled our expertise from developing cores for data centers to meet these automotive performance demands."
Arm’s new lineup of AE processors
Augmenting the AE product range are the Arm Cortex-A720AE and Cortex-A520AE processors. The Cortex-A720AE is meticulously engineered with automotive needs in mind, providing consistent high performance and versatility in SoC design. It is versatile enough to handle an array of SDV tasks, ranging from infotainment to the intricate requirements of autonomous driving technologies. Conversely, the Cortex-A520AE prioritizes energy conservation, integrating safety functionalities that allow its deployment in diverse automotive use cases without compromising on performance or safety.
Arm showcases the Cortex-R82AE processor as the crown jewel in real-time processing with a focus on functional safety, marking a debut in 64-bit processing for real-time automotive applications, catering to the stringent demands of such systems.
"With the advent of our v9 architecture, developers now have the luxury to craft and refine their software in the cloud's vast computational environment and seamlessly transition their applications onto automotive edge devices, thanks to a unified instruction set architecture," Vachani elaborates. "For software-defined vehicles, this capability is a game-changer, ensuring dependable and instant software deployment."
Completing the AE suite is the groundbreaking Arm Mali-C720AE image signal processor (ISP), adept at managing multiple concurrent pipelines to serve both computer vision and human vision scenarios. Its flexibility in handling diverse visual processing challenges saves memory, conserves power and space on the SoC, and reduces the lead time in image processing workflows. Additionally, developers can adjust the ISP’s processing components to meet specific image processing demands and quality targets, particularly for computer vision-focused functions.
Each AE series processor is also boosted by an array of customizable system IPs, empowering the Arm silicon community to forge automotive SoCs that are not only scalable but also deliver exceptional performance.
Computing Paradigms for Software-Defined Vehicles
The traditional linear development paradigm, where hardware conception and production lead the way before any software comes into play, is becoming a bottleneck in the dawn of software-defined vehicles. This approach prolongs the time it takes to bring products to market, hinders the rollout of innovative features, and can erode an automaker's competitive standing. With an increasing dependence on software for essential operations and distinctive functionalities, Original Equipment Manufacturers (OEMs) are compelled to adopt more agile methods to swiftly iterate and propagate software enhancements.
"The age-old sequence of crafting hardware, awaiting market arrival of the silicon, and then commencing with software development is no longer feasible for automotive needs—it's simply obsolete," Vachani asserts. "For a meaningful reduction in time-to-market for automakers, there must be a concurrent launch of both hardware and corresponding software enablement platforms right from the outset."
Moreover, the computing power demanded by artificial intelligence and machine learning algorithms eclipses that of conventional automotive computing duties by several magnitudes. Such algorithms necessitate not just high-caliber computing cores but also tailor-made processors capable of adeptly managing concurrent operations, neural network computations, and immediate data analysis, all while adhering to the strict energy consumption and thermal limits that are characteristic of automotive settings.
Example of an automotive compute subsystem
Confronting these modern-day dilemmas calls for a new breed of automotive processors—ones equipped with the prowess to deliver the requisite computational performance and judicious use of power. This entails processors that embrace avant-garde architectures, refined for swift and effective handling of AI-driven tasks, fortified security protocols to guard against the ever-advancing landscape of cyber threats, and the adaptability to cater to a sweeping array of automotive functionalities, from advanced driver-assistance systems (ADAS) to the zenith of self-driving technology. In parallel, the advent of virtual prototyping and software-centric development frameworks bestow upon developers the boon of initiating software conception and validation stages well in advance of their hardware counterparts, drastically truncating the development timeline and fostering more rapid deployment of novel technologies to the marketplace.
"Our unveiling of virtual prototyping tools stands as a singular milestone, setting this launch apart from anything previously undertaken with the Arm ecosystem," proclaimed Vachani.
“Think about the digital twin technology that's in the factory automation space. We're essentially bringing this concept to the automotive market. This is what allows the automotive developers to work together early ahead of physical design being available.”
Arm Elevates Automotive Computing to New Heights
To excel in the burgeoning landscape of software-focused vehicular design, car manufacturers are urged to cultivate an integrated development ethos, merging state-of-the-art processing technologies with pioneering software engineering practices. Arm's latest ensemble of AE processors aims to equip car engineers with the computational tools and developmental backing essential for prosperity in this new domain.
"This suite of solutions lays the groundwork for bolstering our ecosystem with an all-encompassing chip-system-software (CSS) offering, propelling them towards swifter, more nimble, and cost-efficient operations," stated Vachani. "It includes ready-to-deploy, production-grade RTL subsystems, meticulously tailored for the safety requisites and software exigencies of vehicular systems. This enables our partners and ecosystem to channel their precious resources into areas of unique innovation, while Arm continues to do what it excels at: forging a scalable and efficacious compute infrastructure."
