Embracing the Future with Cortex-A320: A Deep Dive into the General Armv9 Architecture Adoption

By Christophe Fava-Rivi, Director, Software Product Management, IoT Line of Business

February 27, 2025

Discover how the Arm Cortex-A320, the first ultra-efficient Armv9 CPU, brings advanced Armv9 features and benefits to IoT markets.

The introduction of the Arm Cortex-A320 CPU marks an important milestone: It’s the first ultra-efficient implementation of the Armv9 architecture. This groundbreaking CPU brings advanced capabilities previously reserved for leading-edge mobile computing solutions to power-constrained devices, delivering significant improvements in AI processing, security and overall efficiency.  

Sounds good so far. But why bring Armv9 capabilities to bear in a processor technology intended to serve incredibly diverse edge and endpoint devices already well served by other Arm processor implementations? Because it’s time.   

Today’s IoT landscape demands more from edge devices than ever. Smart cameras need to run sophisticated computer vision algorithms locally. Industrial sensors must process complex machine learning (ML) models for predictive maintenance. Even simple endpoints increasingly require enhanced security and virtualization capabilities. These evolving demands make Armv9’s advanced features no longer just nice-to-have, but essential for the next generation of IoT innovation. 

The Armv9 architecture brings to the edge the transformative capabilities of Cortex-A320 by leveraging key Armv9 features including Scalable Vector Extension 2 (SVE2) for enhanced AI and digital signal processing, comprehensive security features like Memory Tagging Extension (MTE) and advanced virtualization support through Secure EL2 (S-EL2). These capabilities, combined with Cortex-A320’s efficient microarchitecture, create new possibilities for AI processing at the edge while maintaining strict power budgets. 

Let’s explore some of these capabilities. 

The power of general Armv9 Architecture adoption 

A standout feature of the Armv9 architecture is its support for SVE2. SVE2 provides significant improvements in digital signal processing (DSP) tasks, SVE2 enables faster and more efficient processing of complex algorithms. This is particularly beneficial for applications that require high computational power, such as AI and ML workloads. With the aid of SVE2, smart cameras can process video streams more efficiently, voice interfaces can handle natural language processing (NLP) with lower latency and industrial sensors can run sophisticated analysis algorithms while maintaining long battery life.  

See also: 

Accelerating Video Decode And Image Processing With Armv9 CPUs And SVE2 

Learn the architecture – Introducing SVE2 guide 

Advanced Security of Cortex-A320

In today’s digital age, security is paramount. Cortex-A320 addresses this need with advanced security features, including MTE, Pointer Authentication (PAC), and Branch Target Identification (BTI). These features work together to help safeguard against various cyber threats. 

MTE helps detect and mitigate memory safety vulnerabilities, which are common in C/C++ programs. By tagging memory allocations and checking these tags during access, MTE can identify and prevent potential security breaches. PAC adds an extra layer of security by ensuring the integrity of function pointers and return addresses, making it harder for attackers to exploit software vulnerabilities. BTI, on the other hand, protects against control flow attacks by ensuring that indirect branches only target valid locations. 

See also: 

Enhanced security through Memory Tagging Extension 

Enabling PAC And BTI On AArch64 For Linux 

Part 2: Enabling PAC And BTI On AArch64 For Linux 

Part 3 Enabling PAC And BTI On AArch64 For Linux 

Learn the architecture – Providing protection for complex software 

S-EL2 Virtualization for enhanced isolation

Virtualization is a key technology in modern computing, enabling the efficient use of resources and improved isolation between different workloads. Cortex-A320 supports S-EL2 virtualization, which enhances the isolation of virtual machines (VMs) running on the same hardware. This is particularly important in multi-tenant environments, where different users or applications share the same physical resources. 

S-EL2 provides a secure execution environment for VMs, ensuring that sensitive data and operations are protected from other VMs and the underlying hypervisor. This level of isolation is crucial for maintaining the integrity and confidentiality of data in cloud and edge computing scenarios. 

To keep sensitive data and code safe from unauthorized access, TrustZone is a built-in security feature in Arm processors that creates a separate, protected environment. This ensures that critical tasks run in isolation from potential threats. Alternatively, Hafnium is a secure firmware reference implementation for A-class Arm processors, offering a robust foundation for trusted applications and enhancing system security against cyber threats. 

See also: 

Learn the architecture – TrustZone for AArch64 

Hafnium Documentation 

DSP uplift with SVE2 

SVE2 plays a crucial role in enhancing the DSP capabilities of Cortex-A320. DSP tasks are essential in various applications, including audio and video processing, telecommunications, and scientific computing. SVE2 extends the capabilities of the Armv9 architecture by providing new instructions and data types that optimize the performance of these tasks. 

For instance, SVE2 introduces new instructions for matrix multiplication, which is a fundamental operation in many ML algorithms. These instructions enable faster and more efficient processing of large datasets, leading to improved performance in ML workloads. Additionally, SVE2 supports new data types, such as BFloat16, which are optimized for ML and AI applications. 

See also: 

Arm releases SVE2 and TME for A-profile architecture 

Leveraging the vast Armv9 software ecosystem

One of the key advantages of Cortex-A320 is its compatibility with the extensive Armv9 software ecosystem. This ecosystem includes a wide range of tools, libraries, and frameworks that have been developed and optimized for next-gen edge AI starting from highly optimized compilers (LLVM with loop optimization, cryptography and Single Instruction Multiple Data (SIMD)). By leveraging this ecosystem, our more than 20 million developers can take advantage of the latest advancements in software technology and accelerate the development of their applications. 

The Armv9 software ecosystem includes support for popular operating systems, such as Linux and Android, providing enhanced performance and security features as well as containerization and cloud development methods. This ensures that developers have access to a wide range of tools and resources to build and deploy their applications efficiently. 

See also: 

Arm Toolchain For Embedded: Next-Generation Arm C/C++ Embedded Compiler 

Extending LLVM Optimization 

Running real-time OSes 

Cortex-A320’s compatibility with real time operating systems (RTOS), such as Zephyr and others, enhances its versatility for IoT and embedded applications. Zephyr, a scalable RTOS designed for resource-constrained devices, supports diverse hardware architectures and communication protocols, enabling efficient and reliable development. 

Kleidi: AI performance at the edge

Arm KleidiAI, a lightweight, open-source AI library, optimizes and accelerates AI workloads on Cortex-A320, thanks to key ML framework and runtime integrations. This lets developers leverage the advanced capabilities and flexibility of the Armv9 architecture. Kleidi unlocks AI acceleration on Arm CPUs, optimizing software-level performance across diverse workloads. Its highly optimized kernels enhance leading AI frameworks like ExecuTorch and LiteRT (formerly Tensorflow Lite), enabling faster edge AI execution and seamless workload flexibility between CPUs and NPUs.  

Now, Arm Kleidi is expanding to IoT, unlocking CPU performance for next-gen edge AI applications. Delivering significant acceleration across embedded and IoT use cases, Kleidi boosts Cortex-A320 performance by nearly 70% when running Microsoft’s Tiny Stories small language model (SLM) on Llama.cpp. This powerful combination simplifies AI development and accelerates performance in billions of devices, making it easier for developers to execute the right AI workload in the right place, at the right time. 

See also: 

Kleidi – Software-Level AI Acceleration 

KleidiAI: Helping AI Frameworks Elevate Their Performance on Arm CPUs 

Cortex-A320 is shaping the future of IoT 

The launch of the Cortex-A320 CPU marks a significant milestone in the evolution of computing technology. With its advanced features and robust architecture, Cortex-A320 is set to revolutionize the IoT industry and pave the way for new and innovative applications. The general Armv9 architecture adoption, enhanced security features, S-EL2 virtualization, DSP uplift with SVE2, and compatibility with the vast Armv9 software ecosystem, running RTOS like Zephyr, and compatibility with Arm Kleidi make Cortex-A320 a compelling choice for developers and businesses alike. 

As we move forward, bringing Armv9’s capabilities to ultra-efficient devices opens entirely new possibilities for innovation at the edge. Developers can now imagine and create applications that were previously impossible in power-constrained environments: smart cameras that run sophisticated AI locally, industrial systems that process complex machine learning models in real-time, and IoT devices that maintain enterprise-grade security without compromising battery life. 

Whether you’re developing IoT solutions, AI and ML applications, or secure computing environments, the Cortex-A320 provides the tools and capabilities you need to succeed in the AI age.

Learn more about the Cortex-A320 and its transformative impact on IoT with unparalleled performance, security, and energy efficiency.