Source: Arm Author: Arm
News highlights:
    The world's first Armv9 edge AI computing platform, built on a Cortex-A320 CPU and Ethos-U85 NPU, is optimized for iot applications and supports end-to-end AI models running over 1 billion parameters. It is supported by a number of industry leaders including Amazon Cloud Technologies (AWS), Siemens and Renesas Electronics.
    The ultra-energy efficient Arm Cortex-A320 improves the efficiency, performance and security of iot applications through the Armv9 architecture, driving advances in industrial automation, smart cameras and more.
    Extending Arm Kleidi to the Internet of Things can deliver up to 70% performance improvements, enabling more than 20 million developers to seamlessly integrate leading AI frameworks and simplify edge AI development.
    Arm Holdings Limited (NASDAQ: ARM, hereinafter referred to as "Arm") today announced the launch of the Arm? The v9 edge artificial Intelligence (AI) computing platform, which features the new Arm Cortex? -A320 CPU and leading edge AI accelerator Arm Ethos? - The U85 NPU is the core, which can support end-to-side AI models running over 1 billion parameters.

    Paul Williamson, senior vice president and general manager of the Internet of Things Group at Arm, said: "AI innovation is no longer limited to the cloud. As the world becomes increasingly connected and intelligent, from smart cities to industrial automation, handling AI workloads at the edge not only brings significant advantages, but also the necessity. The introduction of the Armv9 edge AI computing platform, built specifically for the Internet of Things, marks an important milestone in this development."

Armv9 Edge AI Computing Platform - Integrate efficiency, security and intelligence at the edge
    Oems are under more pressure than ever to quickly deliver solutions to the rising computing demands of iot applications, including precise navigation of autonomous vehicles in factory environments, smart cameras that flexibly adapt their functions through software upgrades, and building human-machine interfaces that provide more natural AI interaction experiences.
    To innovate and scale quickly, Oems need the flexibility to execute AI workloads in the right place for stronger security and greater software flexibility, and Armv9 technology is ideal to meet these needs in large-scale applications. Arm's computing platform integrates a new, ultra-energy-efficient Armv9 CPU, the Cortex-A320, and the Ethos-U85 NPU with Transformer operator network support, creating the world's first Armv9 edge AI computing platform optimized for the Internet of Things. Compared to the Cortex-M85-based platform launched last year, the new edge AI computing platform offers eight times better machine learning (ML) performance.
    The wide availability of Arm technology enables partners to now deploy powerful Armv9 technology in a variety of computing scenarios, from the cloud to the edge. From partners licensing the technology to build SoC chips, to ODMs and Oems building next-generation devices, the platform is supported by a number of industry-leading partners including Amazon Cloud Technologies (AWS), Siemens, Renesas Electronics, Advantech and Eurotech.

The new Cortex-A320 lays the security foundation for the next generation of smart iot devices
    The launch of the new edge AI computing platform marks an important milestone in the development of edge computing. The Cortex-A320 brings advanced AI capabilities and developer benefits to the Internet of Things, extending the capabilities of the Armv9 architecture to energy efficient devices and providing comprehensive software support.
    The Cortex-A320 takes advantage of the Armv9 architecture, such as SVE2 for ML performance. Compared to its predecessor, the Cortex-A35, the Cortex-A320 delivers a tenfold increase in ML performance and a 30% increase in scalar performance.
    The Armv9.2 architecture of the platform also brings advanced security features to even the smallest Cortex-A devices, such as pointer validation (PAC), branch target recognition (BTI), and memory tag Extension (MTE). These capabilities are critical because edge devices typically operate in exposed environments and process sensitive data.
    With the continued integration of edge AI on the Arm platform, the addition of this new product enables Arm to now offer the most advanced Armv9 Cortex processor family for the Internet of Things, ranging from extreme performance to cost - and energy-constrained devices.

Extend Arm Kleidi to iot

    One of the biggest barriers to edge AI adoption is the complexity of software development and deployment, which is where the Armv9 Edge AI computing platform software ecosystem benefits. Arm has extended Arm Kleidi to the Internet of Things, a set of computing libraries for AI framework developers designed to optimize Arm CPU-based AI and ML workloads with no additional developer action. KleidiAI has been integrated into common iot AI frameworks such as Llama.cpp and ExecuTorch or LiteRT (via XNNPACK), accelerating the performance of key models such as Meta Llama 3 and Phi-3. For example, when running Microsoft's Tiny Stories dataset on Llama.cpp, KleidiAI brought a performance boost of up to 70% to the new Cortex-A320.
    These improvements are critical because in today's fast-paced technology environment, how quickly a product gets to market often determines its success or failure. The new edge AI computing platform ensures seamless software-level compatibility with higher performance Cortex-A processors. This scalability enables developers to create solutions that can be flexibly adapted as requirements change. With the massive Armv9 ecosystem and compatibility with feature-rich operating systems like Linux and real-time operating systems like Zephyr, developers have unprecedented flexibility. They can leverage existing tools and knowledge, as well as the benefits of software reuse, to accelerate time to market and reduce total cost of ownership. There are more than 20 million active Arm developers worldwide, which offers great potential for future technological innovation.

Look to the future

    Looking ahead, it is clear that the future trend of AI will shift to the edge, and the new Arm edge AI computing platform will be the catalyst for a new round of iot innovation. The combination of the features of the Armv9 architecture, advanced AI capabilities and comprehensive software support creates new possibilities for Oems and developers.
    The platform enables the running of tuned Large language models (LLM) and small language models (SLM) on agent-based AI applications, opening up entirely new categories of edge application scenarios. In future scenarios, intelligent decision-making will be closer to the source of data collection, which will not only significantly reduce latency, but also effectively improve the level of privacy protection.
    This is not just an incremental step forward, it represents a fundamental change in how the industry approaches edge computing and AI. This is the first time the industry has welcomed Armv9 cpus optimized for iot applications, combining ultra-high energy efficiency with advanced AI capabilities to achieve unprecedented breakthroughs.

Partner testimonies:
    "The pace of development of edge AI is accelerating, and advances in Arm's iot computing architecture will bring new possibilities to edge intelligence. As a leader in edge computing and edge AI, Advantech believes this innovation represents a significant step forward for the Arm ecosystem, which will help industries create smarter, more efficient, and more secure AI applications. This innovation will spark industry prosperity and technological breakthroughs in the edge computing market."

Zhang Jiahao, President of Advantech Embedded Internet of Things
    "The new Arm edge AI computing platform enables our customers to run the Nucleus Lite, a lightweight device runtime environment that runs AWS IoT Greengrass on Armv9 technology, enabling edge devices to run efficiently with minimal memory requirements. The seamless integration of these two technologies provides developers with optimized solutions to support building modern edge AI applications, such as anomaly detection in precision agriculture, intelligent manufacturing, and autonomous driving."

Yasser Alsaied, head of iot technology at Amazon Cloud Technologies
    "As the iot market continues to expand, Eurotech will continue to provide customers with the performance, security and efficiency solutions required for the latest edge AI applications. Arm's new edge AI computing platform gives us the foundation to build the next generation of diverse iot devices, and Armv9 allows us to take security performance, energy efficiency and software flexibility to a whole new level."

Eurotech Chief Technology Officer Marco Carrer
   "Renesas Electronics is committed to serving the broad AI iot market segment with diverse use cases and increasingly intelligent workloads on a truly scalable computing platform. We are excited about the latest Armv9 Cortex-A320 CPU, which not only delivers outstanding AI/ML performance and security, but also significantly optimizes power consumption and area efficiency. This breakthrough helps us accelerate the pace of innovation and further improve efficiency and scalability."

Daryl Khoo, Vice president of Embedded Processing at Renesas Electronics
    "Siemens is committed to unlocking the potential of AI in edge applications. The new edge AI computing platform based on the Armv9 architecture helps us extend our portfolio of highly secure, high-performance and energy-efficient AI innovations to all customers across a range of industrial, smart infrastructure and mobile applications."
    Herbert Taucher, Vice President, Research and Pre-Development, Integrated Circuits and Electronics, Siemens

About Arm
    As the industry's most powerful and energy-efficient computing platform, Arm covers 100% of the world's connected population with unparalleled scale. Arm delivers advanced solutions to meet the never-ending demand for computing, enabling the world's leading technology companies to unlock unprecedented AI experiences and performance. Arm is working with the world's most extensive computing ecosystem and 20 million software developers to build the future of AI on the Arm platform.