The ARM Cortex-M and Kirin A1 are two very different processors designed for different use cases. The Cortex-M is designed for embedded and IoT applications while the Kirin A1 is designed for wearables and edge AI. Here is an overview comparison of the key differences between these two ARM-based processors.
Intended Use Cases
The Cortex-M processor is designed to be a very efficient and low cost solution for embedded and IoT devices. It excels at basic tasks like reading sensors, running control loops, interfacing with peripherals etc. Cortex-M powered devices may have constraints such as low memory, battery power and cost. Popular applications include industrial automation, robotics, automotive systems, smart home devices and wearables.
In contrast, the Kirin A1 focuses more on wearables and edge AI applications. It has dedicated AI processing capabilities to enable voice recognition, gesture control and facial recognition on devices like smart watches, earbuds and AR/VR glasses. The Kirin A1 is designed to deliver a decent level of AI performance while still maintaining low power suitable for battery-powered devices.
The Cortex-M series utilizes simpler in-order only CPU cores without cache compared to Kirin A1’s out-of-order superscalar cores with cache. Cortex-M cores are based on the ARMv6-M, ARMv7-M or ARMv8-M architectures. Clock speeds typically range from 50 MHz to 400 MHz.
Kirin A1 uses 4 ARM Cortex-A55 cores based on the ARMv8.2 architecture and clocks up to 1.5 GHz. The A55 is relatively power efficient while still delivering better single threaded performance versus the Cortex-M. Cache helps to reduce memory accesses for better efficiency.
Cortex-M MCUs support external memories like SRAM, flash and PSRAM connected via simple bus interfaces. Memory support varies across Cortex-M devices but is generally up to a few MBs on lower end MCUs. Some higher end Cortex-M offers memory interfaces like AXI/AHB bus and can support up to 16 MB RAM.
The Kirin A1 supports LPDDR4X with up to 8GB RAM and UFS 2.1 storage. This is necessary to feed its AI accelerator hardware and also run an operating system like HiSpark OS over Cortex-A55 cores. The extra memory bandwidth helps deliver better overall performance.
Graphics and Display
Basic 2D graphics acceleration may be available via Cortex-M cores interfacing with LCD controllers. This allows simple GUIs, graphics and animations to be rendered on displays. External graphics chips can also be utilized e.g. via SPI port.
The Kirin A1 contains an ARM Mali-G57 GPU that can handle 3D graphics and video decoding up to 4K resolution. This enables smooth visual experiences on up to QHD screens. The A1 also has a HiFi 4 DSP for audio.
Low power machine learning inference is possible on some Cortex-M devices using CMSIS-NN optimized neural network libraries or external AI chips connected over the bus interface. Performance is generally limited to a few hundred MHz and under 1 TOPS.
A key strength of the Kirin A1 is its 1.5 TOPS NPU for ML inferencing. It supports common models like face/speech recognition and keyword spotting while maintaining high efficiency. The NPU combined with the CPU and GPU makes the A1 suitable for wearables requiring edge AI.
Cortex-M MCUs normally include basic connectivity like UART, SPI, I2C, USB, Ethernet, CAN etc. More advanced MCUs may have low power radios for Bluetooth Low Energy, ZigBee, Thread and other mesh protocols. Cellular modems can also be added.
The Kirin A1 has built-in support for Bluetooth 5.2, WiFi 5 and NFC. It can interface with LTE modems and includes a GNSS receiver for location services. Connectivity is vital for wearables interacting with smartphones and other devices.
Secure boot, one-time programmable memory and hardware cryptographic accelerators are common security features on advanced Cortex-M processors. Protocols like TLS can be implemented in software stacks. External security chips can also complement the MCU security.
The Kirin A1 contains dedicated security subsystems for secure storage, secure bootup, trusted execution environments and fast cryptography using the AES and SHA engines. Security is critical for protecting user data and privacy on wearables.
Efficiency is a key design focus for Cortex-M processors. Static and active power is optimized for low power operation under 100 mW. Select MCUs support modes like deep sleep and shut down current under 1 uA. This allows long battery life for energy constrained devices.
Huawei quotes the Kirin A1 power consumption as 20 mW for continuous heart rate monitoring and standby current around 10uA. While higher than Cortex-M, the A1 maintains very low power suitable for smaller wearable batteries.
Operating System Support
The Cortex-M processor primarily runs real-time operating systems like FreeRTOS and ThreadX or bare metal applications. The simplicity of Cortex-M allows developers to program right to the hardware in some cases avoiding complex OS usage.
Kirin A1 supports more full-featured OS environments like HiSpark OS, a RTOS based on FreeRTOS plus Linux distributions like LiteOS. This provides appropriate software infrastructure for smart wearables.
Cortex-M devices are supported by a wide range of development tools including:
- Compilers like GCC ARM, IAR Embedded Workbench
- Debug probes like J-Link, ST-LINK
- IDEs like Keil MDK, STM32CubeIDE
- NXP MCUXpresso, Arduino
Huawei provides the HiHope IDE based on Eclipse for Kirin A1 development. It includes an SDK, debugger, compiler, libraries and documentation tailored for the A1. Their HiWear platform offers additional software services and cloud backend.
The Cortex-M CPU is licensed by ARM but manufactured by various semiconductor vendors into their own SoC products. Popular vendors include STMicro, NXP, Microchip, Renesas, Cypress, Infineon etc. Hundreds of MCU models are available.
Huawei is currently the sole provider of the Kirin A1 chipset. It is manufactured using their advanced 12nm fabrication process. The A1 powers Huawei’s own wearable products like the Watch GT2 Pro.
Affordable cost is a major appeal of Cortex-M processors. Basic MCUs start below $1 while higher end devices range up to $10 or $20. High volume production can bring the cost down significantly. The simple architecture also allows very small MCU packages.
As a newer chip targeted at flagship wearables, the Kirin A1 carries a higher cost estimated around $20 per unit. But this enables the advanced CPU, AI, connectivity and security necessary for next-gen wearables.
In summary, the Cortex-M processor excels at low cost, low power embedded applications with basic MCU requirements. Its simplicity, ubiquity and software maturity are key strengths.
Kirin A1 differentiates by targeting wearables and edge AI capabilities. Its CPU, graphics, ML hardware are designed to offer advanced on-device computing. Connectivity and OS support enable rich user experiences and applications.
Each processor suits different segments of the ARM-powered device ecosystem. Cortex-M will continue dominating basic embedded devices while Kirin A1 brings new possibilities for wearables and on-the-go AI.