Why Cortex-M85?

The spread of IoT and edge AI has led to an era in which microcontrollers are required to have higher computing power than ever before. To process the massive amounts of data acquired from sensors and cameras in real time and perform AI inference, a simple control MCU does not have enough performance.
Conventional processors such as the Cortex-M7 have limitations when implementing AI or advanced DSP processing, making it necessary to use a higher-performance MPU. However, MPUs often have disadvantages in terms of power consumption and cost, making their application to embedded devices problematic.

Against this background, Arm has introduced the Cortex-M85 as a new architecture that can speed up AI and signal processing while maintaining the low power consumption of MCUs.

What is Cortex-M85?

The Arm® Cortex®-M85 is a high-performance microcontroller core that uses the Armv8.1-M architecture. Its main features are as follows:

Industry-leading performance: Achieves 6.39 CoreMark/MHz, delivering 4x​ ​the ML performance and 3x​ ​the DSP performance of the Cortex-M7.

- Helium™ MVESupport: M-Profile Vector Extension (MVE) accelerates DSP and machine learning (ML) tasks.

Advanced security: Secure system isolation and attack protection with TrustZone® and PACBTI.

- Low power consumption design: Achieves processing power on par with that of an MPU while maintaining the power-saving features unique to MCUs.

This makes it possible to implement AI applications such as image recognition and voice processing, which previously required an MPU, using a low-cost, low-power MCU.
 
Below is a comparison of the Cortex-M series cores: Cortex-M7, Cortex-M55, and Cortex-M85.

Comparison of Cortex-M7 / M55 / M85

	Cortex-M7	Cortex-M55	Cortex-M85
architecture	Arm v7-M	Arm v8.1-M	Arm v8.1-M
Security	―	―	PACBTI
		Unprivileged Debugging Extensions	Unprivileged Debugging Extensions
		Stack Limit Check	Stack Limit Check
		TrustZone	TrustZone
	MPU (PMSAv7)	MPU（PMSAv8）	MPU (PMSAv8)
pipeline	Six-stage superscalar and branch prediction	4 stages (for the main integer pipeline)	7-stage scalar pipeline and 9-10 stage vector and floating-point pipeline
Helium（MVE）	Not supported	Supported	Supported
FPU	fp32, fp64	fp16, fp32, fp64	fp16, fp32, fp64
	FPv5	FPv5	FPv5
MACs per cycle	1 32bx32b	8 8bx8b	8 8bx8b
		4 16bx16b	4 16bx16b
		2 32bx32b	2 32bx32b
CoreMarks/MHz	5.29	4.4	6.28
DMIPS/MHz	2.31/3.23/6.78	1.69/2.16/5.32	3.13/4.52/8.76

Figure 1. CM85 vs CM7 and CM55 benchmarks [Data source: Arm]

The appeal of Helium (MVE)

Helium (MVE) is a system that can process many calculations at once with a single command.

The conventional calculation method was like "one person processing one piece of data in sequence," butcalculation usingHelium(MVE) is like "a team of fourpeople performing the same task simultaneously."

The diagram belowshows the difference between the conventional calculation method and the calculation method used byHelium(MVE). Conventional calculations are scalar processing, wherecalculations are performedone by one in order, but this method takes a long time to process. Helium(MVE)uses theSIMD method, which processes multiple data in parallel with a singleinstruction. This allows calculations to be completed in a short time, greatly improving processing efficiency.

Figure 2. Comparison of conventional calculation methods and Helium (MVE) calculations

By utilizing Helium (MVE), multiplication and addition operations that are commonly used in AI learning and inference can be accelerated, resulting in the following benefits:

- Accelerated AI inference: Improved efficiency of TinyML tasks such as image classification and speech recognition.

- Enhanced DSP processing: Faster filtering and signal processing.

- No additional DSP required: High-load calculations can be performed using the MCU alone, reducing system costs.

For example, in image recognition, it is possible to simultaneously calculate "pixel value x filter coefficient," and in audio processing, it is possible to simultaneously calculate "signal x filter coefficient."

The figure below compares DSP processing performance with and without Helium (MVE) in Arm Compiler AC6.16.
The horizontal axis represents "normalized performance," which is calculated using the performance without MVE as the reference (1.0), and shows how much faster processing performance is achieved when MVE is used. As a result, we can see a performance improvement of approximately 57 % for CFFT-32 and approximately 64 % for FIR-32. In other words, using Helium makes signal processing such as FFT and FIR more efficient.

Figure 3. Improved DSP processing performance with Helium (MVE) [Data source: Arm]

RA8E1 Series Overview

The Renesas RA8E1 is an entry-level model equipped with the Cortex-M85 and has the following features:

- Operating frequency: Maximum 360MHz

Memory configuration: 1MB flash, 544KB SRAM (including TCM)

・Main peripheral functions: Ethernet, CAN-FD, USB, 8-bit camera interface (CEU)

The RA8E1 series is the most cost-effective product in the RA8 family. The evaluation board FPB-RA8E1 is offered at an affordable price, making it ideal for those who want to try the Cortex-M85 for the first time.

First-time setup procedure

We will introduce the RA8E1 development process, which is easy even for beginners to get started with.

Step 1: Get an evaluation board

Purchase the FPB-RA8E1 evaluation board from the Renesas official website. It can be powered by USB, so it's safe to use in a home environment.

FPB-RA8E1 - RA8E1 Fast Prototyping Board | Renesas

Summary of what you can do with Cortex-M85

The Cortex-M85 combines high performance and power efficiency that exceeds conventional standards for microcontrollers. AI It is a reliable option for achieving high-speed signal processing at the edge.
RA8E1 The group provides a development environment that is easy to use even for beginners and a wealth of samples, making "high-performance microcontroller development that anyone can do" even more accessible.

Reference links:

​Unprecedented DSP and AI performance in a microcontroller core leveraging Helium and ARM® Cortex®-M85 | Renesas

​RA8E1 - 360 MHz Arm Cortex-M85-based entry-level microcontroller with Helium and TrustZone | Renesas

​RA8E1 - 360 MHz Arm Cortex-M85-based entry-level microcontroller with Helium and TrustZone | Renesas

Contact Us