Arm Cortex-A320: The Smallest Armv9 Implementation for Edge AI Applications

The Arm Cortex-A320 is a low-power Armv9 CPU core optimized for Edge AI and IoT applications, offering up to 50% power reduction over the Cortex-A520. It’s the smallest Armv9 core to date. Introduced in 2021, Armv9 has been used in flagship mobile devices, smartphones, and high-end motherboards. The upcoming Rockchip RK3688 AIoT SoC also features Armv9 but targets high-end applications. The Cortex-A320 expands Armv9 to a wider range of IoT devices, including power-constrained Edge AI devices.

The Cortex-A320 features several key attributes that fulfill the demands of edge AI applications. Its design offers over 50% power reduction compared to the Cortex-A520, making it a great choice for power-constrained devices. The processor also delivers enhanced AI processing capabilities, with up to 10x ML processing uplift compared to the Cortex-A35. Additionally, it provides robust security features, including Memory Tagging Extension (MTE), Pointer Authentication (PAC), and Branch Target Identification (BTI).

Specifications of the Arm Cortex-A320

• Architecture
  • Arm9.2-A (Harvard)
• Extensions
  • Up to Armv8.7 extensions
  • QARMA3 extensions
  • SVE2 extensions
  • Memory Tagging Extensions (MTE) (including Asymmetric MTE)
  • Cryptography extensions
  • RAS extensions
• Microarchitecture
  • Up to 4x CPUs in cluster
  • 40-bit Physical Addressing (PA)
  • In-order pipeline
  • Partial superscalar support
  • NEON/Floating Point Unit
  • Optional Cryptography Unit
• Memory system and interface
  • 32KB or 64KB L1 I-Cache / D-Cache
  • Optional L2 Cache – 128KB, 192KB, 256KB, 384KB, or 512KB
  • No L3 Cache
  • ECC Support
  • Bus interfaces – AMBA AXI5
  • No ACP, No Peripheral Port
• Security
  • Trust Zone, Secure EL2, MTE, PCA/BTI

Features and Applications

The Arm Cortex-A320 features an optimized microarchitecture, with a single-issue, in-order CPU and a 32-bit instruction fetch. Its 8-stage pipeline allows for higher frequency points, while the forwarding network reduces power consumption. Also, the processor supports up to 64KB L1 caches and up to 512KB L2 caches, providing fast memory access. Its 256-bit AMBA5 AXI interface supports high-bandwidth memory access, making it suitable for demanding edge AI applications.

The Cortex-A320 offers several features and optimizations that set it apart from other processors. Its design and optimized microarchitecture provide significant improvements in performance and power consumption. The processor’s support for Armv9 architecture, including NEON and SVE2 vector processing technologies, brings advanced ML/AI applications. Additionally, its security features like Secure EL2, MTE, PCA/BTI, and TrustZone with scalable design make it a reliable processor for a wide range of applications.

Software Support for the Cortex-A320

A comprehensive software ecosystem, including extensive open-source Linux support and a strong security ecosystem support the Cortex-A320. Its A-profile architecture supports software portability for Android and other rich operating systems. The processor suits various applications, including edge AI, IoT, smart wearables, and infrastructure devices. Its scalability and versatility make it a strong contender for targeting multiple markets in IoT and beyond.

Renesas may be the first to launch an Arm Cortex-A320 SoC in 2026, as one of Arm’s key partners. You can find more details about the product on the Arm newsroom and Arm’s developer website.

Please follow and like us: