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

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. 
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 
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 
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 
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 
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 
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. 
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 
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.
Any re-use permitted for informational and non-commercial or personal use only.
A new generation of AI-native companies is emerging, bringing innovations that could completely disrupt industries as we know them.

Speaking with @theCUBE, Arm CMO Ami Badani shares how Arm’s compute platform is driving this + her advice on staying ahead: 
Ami Badani, Arm | theCUBE + NYSE Wired: CMO Leaders Summit
Arm Badani at Arm talks with John Furrier at theCUBE + NYSE Wired: CMO Leaders Summit.
okt.to
We’ve worked with OpenCV on the first ever KleidiCV integration! 🎉 
 
With 4x performance uplifts on OpenCV 4.11, in Android KleidiCV is enabling faster responses across key computer vision tasks on mobile with no extra developer effort. 
 
Get the details:…
📢 Discover a new way to enhance your code with the Arm extension for @GitHub Copilot.

Available now on GitHub Marketplace, it makes migration to Arm simple with tailored architecture-specific answers and performance optimization tips: https://okt.to/CWdnEy
🤝 Together with @Synopsys we’re accelerating time-to-market for Arm CSS-based designs.
 
Earlier this week, Kevork Kechichian joined the team to celebrate the new ZeBu-200 and HAPS-200 hardware platforms that speed up time-to-market for AI innovations: https://okt.to/5WmCVL
We’re heading to #SXSW 2025! 🚀 

Join Arm CEO Rene Haas on Wednesday, March 12, as he shares insights on the AI era and the unprecedented opportunities being unlocked on the Arm compute platform.

Add it to your agenda: https://okt.to/609gAn
🥳 Congratulations to our new group of Distinguished Arm Ambassadors who lead and inspire developers every day.

Want to be a part of the community?

🔗 Applications for the Arm Ambassador Program are now open, join us!
Here’s to our new Distinguished Arm Ambassadors. 👏

Their continued support and exceptional contributions to the Arm Developer Program help inspire the software community to learn and innovate on the Arm compute platform. 

Thank you for everything you do!
Think chips are just in your gadgets? Think again. 🧠

@crmiller1 joins Rene Haas on Tech Unheard to reveal how semiconductors are powering our world – from tech innovation to globalization.

🎧 Search Tech Unheard or listen here: https://okt.to/o93Jtf
It’s been another record-breaking quarter in Q3 FYE25!

With revenue at an all-time high, we’re continuing to see strong momentum across the Arm ecosystem, with ongoing adoption of high-performance, power-efficient Armv9 and CSS platforms driven by AI:
Low power is more critical than ever in automotive.⚡

Catch Dipti Vachani’s interview with @TheSixFiveMedia’s @OABlanchard as they discuss what’s next for automotive innovation, and how low power is not just in Arm’s solutions but a core part of our culture and foundation.
Dipti Vachani, @Arm’s Automotive GM, joins analyst @OABlanchard to reveal how Arm is fueling the future of AI in cars. 🚗🌬️ Watch the full interview for her insights on the next generation of smart vehicles at CES 2025! https://youtu.be/8P9Cb8e8kOw
#CES2025 #OnArm #AI #automotive…
Together with @OpenAI, @SoftBank_Group Corp, & SoftBank Corp, we’re bringing AI agents to work.
 
Through Cristal intelligence, we’re developing Advanced Enterprise AI built on the Arm compute platform, setting new standards in productivity & efficiency: https://okt.to/oEc6CV
📱 #ICYMI: The new #GalaxyS25 Series of AI smartphones are coming from @SamsungMobile!
 
Powered by the Arm compute platform, the new devices bring next-gen AI capabilities to Android as part of a shared vision with Google.
 
Congrats, Samsung! #GalaxyAI

Galaxy S25 AI Phone Range Features Explained | Samsung UK
Learn about all the incredible features on the new Samsung Galaxy S25 AI Phone range at Samsung UK.
okt.to
With the first public spec of the CSA now available, Eddie Ramirez shares how it came about and what it means for the industry.

Speaking at the Chiplet Summit last week, Eddie shares why the CSA will help designers connect chiplets into composable SoCs:.
Arm Defines a New Standard for Chiplet Interoperability
In this video, learn why Arm is investing in chiplets as a way to build dynamic options for compute in the A…
okt.to
📸 Here’s a glimpse of an incredible week with Arm at #wef25 in Davos! ➡️

Explore our blog post for key insights from our panel discussion with world-leading AI academics: https://okt.to/rsdH5V
We’re driving the future of automotive. 🚙

Watch Dipti Vanchani’s insightful talk with @TheSixFiveMedia’s @OABlanchard as they discuss industry trends and Arm’s role in automotive AI innovation.
🚗 @OABlanchard and @Arm’s Dipti Vachani discuss 3 big trends in automotive AI: electrification, autonomy, & redefining the user experience. 🚘 Catch the full interview to see what’s driving the future of smart cars. https://sixfivemedia.com/events/ces-2025/how-arm-is-powering-the-next-generation-of-ai-enabled-vehicles/ #CES2025 #OnArm #AI #automotive…
Rene Haas spoke to @SquawkStreet about The Stargate Project and Arm’s involvement as a key initial technology partner.

Discussing its huge potential, Rene shares what Arm-based technology to expect and what he’s most looking forward to about the project: 
Arm Holdings CEO Rene Haas on $500B Stargate project: It's a big, big deal
Arm Holdings CEO Rene Haas joins 'Squawk on the Street' to discuss the $500 billion Stargate AI project, investing…
www.cnbc.com
🏁 The race is on with Android gaming on the new Lenovo Duet, built on the Arm-based @MediaTek Kompanio 838.
 
With powerful octa-core processing, enhanced multimedia experiences and all-day battery life, the race is in your hands – all you need to do is win!
Chances are, you’re hearing a lot more about semiconductors now than you once did.

🎙️ For the latest Tech Unheard episode, Economic Historian @crmiller1 joins Rene Haas to explore how AI is driving this change, and what it means for the future.

Full ep: https://okt.to/yHqsjm
To unlock the full potential of AI, we need tight collaboration between industry & academia to equip the next-generation with the skills to thrive.

Arm CMO Ami Badani opened up this important conversation at #wef25 with academic experts in the field of AI https://okt.to/Wp5j4q
With new AI applications emerging daily, Arm CMO Ami Badani shares why Arm is the trusted platform to meet the insatiable demand for compute.

Catch the full interview at the link below 👇
We’re proud to have been recognized as a winner of Glassdoor’s Best Places to Work in 2025, ranking number 1 on the UK Large list! 🏆🥳

The award, based on employee feedback, underscores our commitment to our core beliefs and helping our people to do their best work.
We’re excited to be a key initial technology partner in Stargate, bringing the power and pervasiveness of Arm’s compute platform to this bold initiative which will unlock AI’s potential for all.

This is how we shape the future of AI. 
Announcing The Stargate Project | SoftBank Group Corp.
The Stargate Project is a new company which intends to invest $500 billion over the next four years building new…
okt.to
🆕 The first public spec of the Chiplet System Architecture is now available!

This marks a crucial step toward a shared understanding of how to define & connect chiplets into composable SoCs that address varying AI workloads & different industry needs: https://okt.to/lqgceB
Lasting AI innovations must deliver more compute with the same level of energy.

We’re committed to addressing this challenge head-on.

For @TIME, Rene Haas reflects on Arm’s journey, from phones to the data center, that’s paved the way for this moment: 
Rene Haas on What Makes Arm's Chips Indispensable
The CEO talks about Arm’s role in the Fourth Industrial Revolution, the speed of AI advancement and sustainability efforts
okt.to
We’re in Davos for an insightful week at #wef25! 🇨🇭

Throughout the week, we’ll be sharing more about our collaboration across industries to unlock the new experiences and capabilities that AI promises on the Arm compute platform: https://okt.to/UcIvSf

source

About The Author