As manufacturers begin to release high-performance MCUs, it is becoming increasingly difficult to distinguish between MCUs and MPUs.
Back in 2022, when Arm launched the Cortex-M85, an engineer remarked, “Although the M85 is a Cortex-M, I believe its performance has surpassed that of early Cortex-A processors. The M85 processors are no longer just microcontrollers; they can be considered micro-application processors.”
Recently, manufacturers have started to invest more in M85 MCUs, and these products seem to truly possess capabilities that rival those of MPUs. What are the highlights of these products? EEWorld will summarize today.
Understanding the M85’s Inner Workings
Performance is one of the most straightforward advantages of the M85.
The M85 is the first to support over 6 CoreMarks/MHz and over 3 DMIPS/MHz. At the same time, through Arm Helium technology, compared to its predecessor M7, the M85 achieves a 4-fold improvement in DSP and ML processing.This scalar performance level benefits from many innovative features at the microarchitecture level, including optimized dual-issue and selective triple-issue capabilities, improved branch prediction, and an enhanced memory system that includes data prefetching.
According to Arm, in the face of increasingly demanding computational requirements, developers of Cortex-M microcontroller systems face a choice: optimize software to extract more clock cycle processing power from existing microcontrollers or migrate codebases to higher-performance microprocessor architectures? Cortex-M microcontrollers have multiple advantages, including deterministic execution, short interrupt latency, and advanced low-power management modes. Transitioning to Cortex-A processors means sacrificing some of the inherent advantages of Cortex-M. Therefore, Cortex-M needs to launch a product with such explosive performance.
In simple terms, it is about satisfying the desire for both high performance comparable to MPUs and retaining the low BOM cost control, low power consumption, and ease of development associated with MCUs.
The memory system architecture is another advantage of the M85.
The M85 is equipped with a tightly coupled memory (TCM) low-latency memory system that ensures deterministic operation. This architecture provides SoC designers with four 32-bit data TCM interfaces and one 64-bit instruction TCM interface, all of which integrate error correction code (ECC) functionality. The new 32-bit AHB access interface port allows external DMA controllers to concurrently access the M85 processor core’s TCM, supporting various common data stream processing application scenarios.
Scan the code and send the secret code: MCU, to join the relevant technical exchange group
The M85 memory system also includes a level 1 cache system equipped with ECC, connected to external memory via the AMBA 5 AXI master interface, maintaining performance optimization even in the face of low-speed non-deterministic memory access.
Security is another major selling point of the M85.
The M85 introduces the Armv8-M architecture’s TrustZone technology and becomes the first Cortex-M processor to integrate the Armv8.1-M pointer authentication and branch target identification extension (PACBTI), significantly lowering the threshold for developers to achieve PSA security certification level 2.PACBTI provides enhanced protection against return-oriented and jump-oriented software attacks by validating function calls and return addresses.
In summary, the unprecedented performance level brought by the M85 opens up new horizons for MCU developers, and manufacturers have expressed their readiness to follow suit. However, updates and iterations do not happen overnight, and how the M85 is positioned will also be a question. Now, perhaps it is time for the M85 to explode onto the scene.
ST Takes Action
Recently, ST launched the STM32V8, sparking discussions among many users on EEWorld. This product is one of the highest-performing in the STM32 series of microprocessors, achieving a terrifying EEMBC CoreMark score of 5072, significantly higher than that of the STM32N6.
18nm, 800MHz Cortex-M85, 4MB PCM (phase change memory), MVE (Arm Helium M-profile Vector Extension) — these keywords all reflect the strength of this product.
According to ST, simply upgrading to the M85 core improves scalar mathematical operation performance in real-time control applications by 20%; compared to Cortex-M7 products, performance improves by 3.5 times at the same clock frequency; applications relying on DSP functions see performance increases of up to 300%~400%.
The reason for such powerful performance is closely related to its use of the 18nm FD-SOI PCM process, which is the same as that used in the automotive MCU Stellar series. FD-SOI forms a competitive landscape with FinFET, which also provides high-performance computing. This technology is relatively new in the MCU field, enabling the STM32V8 to achieve higher operating speeds while significantly optimizing power efficiency.
PCM as an embedded non-volatile memory (NVM) technology has developed various forms, including Intel’s discontinued 3D XPoint memory, known for its speed. Due to the limitations of embedded flash (eFlash) technology, MCUs have been stuck at the 40nm process node. To shrink the process, the industry has turned its attention to new storage technologies.
PCM is a storage technology that competes with MRAM, FRAM, and RRAM. PCM offers the smallest storage units in its class, allowing for more than double the information storage capacity in the same area, providing physical advantages and achieving the smallest storage area. It can double the overall storage capacity without increasing costs, undoubtedly leading to further cost reductions.
More importantly, compared to MRAM and FRAM, ST’s PCM has additional advantages: it supports a maximum operating temperature of 140°C, and its radiation resistance makes it suitable for aerospace applications as well as meeting the stringent requirements of industrial and automotive applications. Therefore, we see that a significant application of the STM32V8 is in aerospace.
This series of chips supports up to 4MB of ePCM and 1.5MB RAM, with external serial memory interfaces including eight-line SPI (8-bit) and sixteen-line SPI (16-bit), supporting serial NOR flash, HyperRAM, PSRAM, and execute-in-place (XIP) operations.
The significant enhancement of AI capabilities is another advantage of the STM32V8. Through the Arm Helium M-profile vector extension (MVE) technology, it builds the foundation for AI capabilities. Helium, based on packed single instruction multiple data (SIMD) and single-element scalar instructions, supports integer and floating-point operations, covering 8 to 64 bit integer operations and 128 bit vector operations with 16/32 bit floating-point operations. This technology brings enhanced machine learning (ML) and digital signal processing (DSP) capabilities to high-performance microcontrollers, and Arm has packaged MVE into function libraries and code generation tools suitable for C/C++ development.
Along with the new product launch, ST has also upgraded its edge AI product line with an AI model library. This library gathers pre-trained AI models covering various AI application scenarios. These models are optimized for ST’s AI microcontroller product portfolio, accelerating the edge AI development process. According to official information, the model library currently has over 140 models, covering visual, audio, and perception functions required by devices such as wearables, smart cameras and sensors, security devices, and robots.
The STM32V8 also has other upgrades in details, including: integrated Chrom-ART graphics accelerator, TFT LCD controller, and JPEG accelerator; gigabit Ethernet controller supporting time-sensitive networking (TSN), equipped with FD-CAN, high-speed USB, and HDMI-CEC interfaces; includes 12 bit ADC and DAC interfaces, as well as two sets of SAI and four sets of I2S audio interfaces; covers secure storage, secure boot, and secure debugging support; equipped with unique identity and anti-tampering mechanisms, and certified by PSA Level 3 and SESIP 3 security certifications.
Renesas’ Early Layout
Renesas has laid out its plans for the M85 earlier, and its updated products are also exceptionally diligent and frequent.
In October 2023, Renesas launched the industry’s first RA8 series MCU based on the Arm Cortex-M85 processor — RA8M1, achieving a groundbreaking 3000 CoreMark, meeting the requirements for complete determinism, low latency, and real-time operation needed by customer applications. According to Renesas, the performance level of 6.39 CoreMark/MHz will enable system designers to replace commonly used MPUs with RA MCUs.
At that time, Renesas’ first-generation RA8 mainly expanded the product line around the M85 core, using a 480MHz M85 with Helium MVE, and did not introduce new storage types, equipped with 1MB~2MB flash and 1MB SRAM (including TCM).
This June, Renesas quietly launched the “World’s Strongest MCU” RA8P1 series on its official website.
The second generation of the RA8 has upgraded the process to 22nm ULL process, equipped with new storage of 0.5/1MB MRAM (optional 4/8MB flash), and the main frequency has also significantly increased, with the M85 core reaching 1GHz. In comparison, the RT1170 uses a 28nm FD-SOI process technology.
According to Renesas, compared to flash memory, MRAM has faster write speeds, higher durability, and stronger data retention capabilities.Moreover, to pursue better AI performance, the second generation of the RA8 also integrates the Ethos-U55 NPU and another 250MHz M33.
While launching the RA8P1, Renesas also simultaneously launched the RA8D2, which is also equipped with the 1GHz M85 core and new storage MRAM.
This October, Renesas continued to iterate, launching the 1GHz RA8T2 Cortex-M85 microcontroller, integrating MRAM and EtherCAT to empower industrial motor control microcontrollers. This product integrates 1MB MRAM, 2MB with ECC verification SRAM, while providing dual cores with 256KB (M85) and 128KB (M33) tightly coupled memory (TCM), and supports SiP packaging expanded to 8MB of external flash.
Shortly after, Renesas added two new products to the RA8 series, RA8M2 and RA8D2, and refreshed the benchmark, achieving a terrifying 7300 CoreMark.
Renesas stated that the RA8M2 and RA8D2 are equipped with embedded MRAM, which has multiple advantages over flash memory technology — high durability, stronger data retention capabilities, faster write speeds, no erase operations required, byte-addressable, and lower leakage current and manufacturing costs. For higher-demand applications, SIP options with 4 or 8MB of external flash are also available. Additionally, both RA8M2 and RA8D2 MCUs include gigabit Ethernet interfaces and dual-port TSN switches to meet the needs of industrial network application scenarios.
It can be said that Renesas has recently been focusing on high-performance MCUs, and in order to achieve higher processes and stronger performance, it has also been focusing on MRAM.
Final Thoughts
With the introduction of the M85 into the MCU, we are truly struggling to distinguish between MCUs and MPUs. Clearly, these layouts are aimed at AI, which is indeed providing significant momentum for product upgrades.
The M85 is redefining the boundaries of high-performance embedded systems with its extreme performance and heterogeneous integration capabilities, leading MCUs into a new era of “unbounded computing power“.
References
[1]Arm: https://developer.arm.com/community/arm-community-blogs/b/internet-of-things-blog/posts/introducing-cortex-m85
[2]All About Circuits: https://www.allaboutcircuits.com/news/stmicro-unveils-first-18-nm-mcu-for-high-performance-designs/Electronic Design: https://www.electronicdesign.com/technologies/embedded/article/55332065/electronic-design-stmicroelectronics-stm32v8-utilizes-fd-soi-and-phase-change-memory
[4]ST: https://www.st.com/resource/en/product_presentation/stm32v8-presentation.pdf
[5]Renasas: https://www.renesas.cn/
· END ·
Welcome to follow the subscription account of EEWorld: "Robot Development Circle"
Scan the code to add the assistant and reply “robot”
to join the group and exchange experiences face-to-face with electronic engineers.
