The trend in the embedded industry has changed.

Engineers must be well aware of how competitive the embedded industry is. As we enter the second half of the year, manufacturers are still in a fierce race, constantly iterating their products.

Next, EEWorld will take stock of the technological trends in the embedded industry this year.

RISC-V MCU, Rising Rapidly

HiSilicon, a subsidiary of Huawei, recently launched two RISC-V chips, the Hi3066M and Hi3065P. The Hi3066M is an embedded AI MCU designed for the intelligent needs of home appliances. It uses HiSilicon's self-developed RISC-V core, has a built-in eAI engine, and supports a main frequency of 200MHz, 64KB SRAM, and 512KB of built-in Flash. The Hi3065P is a high-performance, large-storage real-time control MCU designed for home appliances, industrial, and other fields. It also uses HiSilicon's self-developed RISC-V core, supports a main frequency of 200MHz, 64KB SRAM, and a maximum of 512KB of built-in Flash, enabling continuous iteration of product functions and algorithm upgrades for customers.

It can be seen that RISC-V has become HiSilicon's next strategic fulcrum, and its "backup plan" may be underway. Recently, HiSilicon also released a series of self-developed chips, including Cat 1 and ADC.

WCH, an established player in the RISC-V field, has always attracted the attention of engineers. With a large number of developers, problems can be quickly resolved. Moreover, its full-stack R & D model of "Qingke RISC-V + Interface PHY" has inherent advantages in promoting the application of RISC-V.

Recently, WCH's dual-core RISC-V MCU, the CH32H417, has been its flagship product. The CH32H417 is an interconnected general-purpose microcontroller based on the dual-core design of Qingke RISC-V5F and RISC-V3F. It integrates a USB 3.2 Gen1 controller and transceiver, a 100Mbps Ethernet MAC and PHY, a SerDes high-speed isolation transceiver, and a Type-C/PD controller and PHY.

Singee's RISC-V MCUs have a good reputation in China. The company's recent flagship product is the high-performance HPM6P00. The flagship product of this series, the HPM6P81, has a built-in RISC-V dual-core with a staggering main frequency of 600 MHz. It supports up to 32 channels of high-resolution PWM output, is equipped with 4 independent 16-bit ADCs (up to 32 analog input channels) and 8 high-speed analog comparators, and integrates high-precision motion control modules such as Σ∆ digital filtering and hardware current loops to meet the requirements of demanding control applications.

The application of RISC-V MCUs in vehicles has also been a hot topic recently. Previously, Infineon announced that it would lead the automotive industry in adopting RISC-V and plans to launch a new series of automotive microcontrollers based on this architecture in the next few years. This series will be included in its well-established AURIX automotive microcontroller brand to expand its existing product portfolio based on TriCore (AURIX TC series) and Arm (TRAVEO series, PSOC series), covering a wide range of automotive applications from entry-level to high-performance, surpassing the current market offerings. In addition, as an important promoter of RISC-V standardization, Infineon is actively deploying the application and ecological construction of RISC-V in the automotive field.

EEWorld has learned that ST is currently observing the opportunities of RISC-V in MCUs and will disclose relevant information to the market in a timely manner if there are relevant plans. In addition, ST is also observing the opportunity to produce automotive MCUs locally in China. The 40nm production technology is completely feasible in China, and ST does not rule out such a possibility.

Automotive MCU, Initiating a Storage Revolution

For automotive MCUs, eNVM is crucial as it is used to store the vehicle's critical code and important configuration data. However, the limitations of eFlash are obvious. For example, the number of rewrites is too small. With each write and erase cycle, the tunnel oxide in the floating-gate NOR cells will degrade, and the leakage current will increase, accelerating the aging of eFlash.

More importantly, eFlash has basically blocked the path of MCU process iteration. Due to the special structure of eFlash transistors, the number of mask layers required for eFlash below 28nm is too large, making it difficult to achieve miniaturization. In addition, the reliability of eFlash beyond 40nm is limited by storage cells, peripheral transistors, and metal interconnects. As the oxide film between transistor devices and metal interconnects becomes thinner, the transient dielectric breakdown (TDDB) lifetime is severely reduced.

To address the above challenges, automotive MCUs are now starting to apply the next-generation eNVM technology, which can be divided into three routes - phase-change memory (PCM), resistive random-access memory (RRAM), and magnetoresistive random-access memory (MRAM).

PCM (Phase-Change Memory): ST recently launched the Stellar series of automotive MCUs with built-in xMemory. Its PCM is based on the 28nm and 18nm FD - SOI (fully depleted silicon - on - insulator) process, and its storage density can be more than twice that of its competitors.

An senior director of ST's automotive MCU division and a member of the division's strategic office explained to EEWorld that the advantages of PCM include five points: 1. PCM provides the smallest storage unit in its class, enabling more than double the amount of information storage in the smallest possible unit area; 2. PCM can double the overall storage capacity without increasing the cost; 3. It has strong high - temperature and radiation resistance and can operate stably even at a junction temperature of 165°C; 4. It can maintain low power consumption under harsh working conditions; 5. PCM is not a new technology, and its technological maturity has been very high since twenty years ago. ST has been researching PCM for many years, so it is very safe and reliable.

MRAM (Magnetoresistive Random - Access Memory): NXP recently launched the world's first 16nm FinFET + MRAM MCU, the S32K5. As a regional controller, it can integrate all real - time control functions and can also be used as a regional aggregator or gateway.

Manuel Alves, senior vice - president and general manager of automotive microcontrollers at NXP Semiconductors, told EEWorld that MRAM has unique advantages. Firstly, its write and programming speeds are extremely fast, ten times faster than flash memory, allowing for quick operation. Secondly, it has strong durability, capable of achieving one million writes. It can not only store code but also be used for data storage, offering high flexibility and facilitating data collection and cross - regional storage.

RRAM (Resistive Random - Access Memory): Infineon launched the AURIX TC4x series of MCUs using TSMC's 28nm process last year and introduced RRAM. It adopts the new - generation TriCore 1.8 architecture, has a main frequency of 500MHz, and is equipped with a PPU parallel processing unit.

Infineon believes that compared with NOR Flash, the RRAM driver is simple. It does not require an erase command before rewriting memory and also supports byte - granularity writing, greatly simplifying the driver design. Fewer commands and memory operations also contribute to the inherent power, performance, and durability advantages of RRAM technology. Compared with EEPROM, RRAM has a faster write speed and higher density, making it suitable for applications that require frequent data updates, such as data recording. Compared with MRAM, RRAM has higher operating power efficiency, offering a balanced combination of performance, power, and cost.

AI MCU, Intensely Competitive

With the development of large AI models, almost every manufacturer now launches MCU products with NPUs and increases investment in software and models.

STM32 is one of the products that every engineer must know. Currently, ST (STMicroelectronics) is increasing its investment in several major trends, including hardware acceleration for edge AI, optimizing the software stack, security functions, and the transition from the edge to the cloud.

In terms of hardware, the STM32N6 is its flagship product. For the NPU, it uses the self - developed Neural - ART accelerator with a frequency of 1GHz, a computing power of 600 GOPS, and an average performance of 3 TOPS/W. For the CPU, the Arm Cortex - M55@800 MHz introduces Arm Helium vector processing technology, bringing digital signal processing (DSP) to the standard CPU. In addition, ST once revealed to EEWorld that it plans to further integrate NPU functions into its Stellar P and G products and will announce detailed plans later.

In terms of software, ST provides Edge AI - Core, Edge AI Developer Cloud, STM32Cube.AI, NanoEdge AI Studio, AI for OpenSTLinux, StellarStudioAI, AIoT Craft, MEMS Studio, MLC/ISPU model library, etc.

Infineon, which topped the global MCU rankings in 2024, also adopts a two - pronged strategy of hardware and software in AI.

In terms of hardware, Infineon's new PSOC Edge E8x MCU series is designed to be the first product to meet the highest certification requirements of the embedded security framework PSA4. All PSOC™ Edge E8x microcontrollers feature on - chip hardware - isolated enclaves with functions such as secure boot, key storage, and encryption operations. Among them, the PSOC Edge E83 and E84 have a built - in Arm Ethos - U55 NPU processor, while the E81 uses Arm Helium DSP technology and Infineon's NNLite neural network (NN) accelerator.

In terms of software, Infineon acquired Imagimob in 2023 and launched the edge AI software solution brand DEEPCRAFT™ in 2024. It can complete the entire process of data collection, pre - processing, model training, optimization, and deployment in one - stop with ModusToolbox. Infineon provides a variety of out - of - the - box models, and the latest models include sound source direction detection models, surface detection models, factory alarm detection models, gesture detection models, fall detection models, etc.

NXP's strategy for edge - side AI focuses on highly integrated hardware, software, and system solutions, emphasizing functional safety, low power consumption, and scalability. NXP believes that Agentic AI has reached the edge, and the autonomous edge is the next step in the industry's development.

In recent years, NXP has increasingly emphasized the concept of solutions and rarely mentions individual products. However, the upgrade of its products has been significant. NXP's eIQ Neutron NPU (Neural Network Processing Unit) has covered three major series: MCUs, crossover MCUs, and application processors. In particular, the crossover MCU i.MX RT700 launched at the end of last year has a huge improvement, which can be described as a "dimensionality - reduction strike." Compared with the previous generation of products, it provides up to 172 times the speed improvement and 119 times the energy savings at the edge, while the power consumption is reduced by 30 - 70% compared with the previous generation of products. The i.MX 9 series of processors also integrate the Neutron NPU, supporting from basic inference to complex multimodal AI applications, including the eIQ AI development kit for optimizing the performance and efficiency of AI models.

Renesas has been advancing rapidly in the MCU field in recent years, with a high product update frequency. For example, it was the first to use cores such as M85 and M55. In the field of edge AI MCUs, the recently launched RA8P1, with extremely powerful performance, combines a 1GHz Arm Cortex - M85 and a 250MHz Cortex - M33 CPU core with an Arm Ethos - U55 NPU, achieving the industry's highest - level 7,300 CoreMark CPU performance and 256 GOPS AI computing performance.