The STM32V8 is the first Cortex-M85 microcontroller with 4 MB of embedded phase-change memory built on the most advanced 18-nm process technology. Running at 800 MHz, it is the first STM32 to exceed 5,000 points in CoreMark, offering a nearly 60% boost compared to STM32H7R/S launched in 2024. The new Arm core also brings essential capabilities, such as its Helium technology (also called M-Profile Vector Extension or MVE), which significantly boosts performance in machine learning and digital signal processing applications. Hence, compared to an STM32H7, an STM32V8 running a computer vision application at 800 MHz can perform inference operations up to six times faster. The STM32V8 will be available during the first quarter of 2026.
Table of Contents
STM32V8: Performance levels enabling next-gen applications thanks to its 18-nm FD-SOI
18-nm
The STM32V8 was made possible thanks in part to ST’s 18-nm process developed in collaboration with Samsung Foundry. Why was the move to 18-nm so significant? Unlike a consumer GPU or CPU, a microcontroller faces the added manufacturing challenge of combining digital and analog circuits, which are much harder to shrink. Unlike in digital logic, a small process node can cause numerous optimization and parasitic problems in analog circuits. Hence, the ability to manufacture an 18-nm STM32 is itself a feat. That’s why ST is the first to feature a Cortex-M85 using this technology. ST will manufacture the new device in our 300-mm fab in Crolles, France, enabling us to control the supply chain better.
FD-SOI
One of the key advantages of our FD-SOI technology is its improved power efficiency, enabling us to run the STM32V8 under high temperature (up to 140°C Tj) to meet the demands of industrial applications. Among many other reasons, this is possible in part thanks to FD-SOI technology, which uses a thin insulator between the source and drain to reduce static leakage, thereby improving efficiency. The increased efficiency also explains why we were able to use embedded phase-change memory (more on that later). Furthermore, the planar process technology used in this process helps make the STM32V8 more cost-competitive.
The Cortex-M85
The Cortex-M85 core itself brings, among other things, an updated Armv8.1-M architecture with a seven-stage pipeline, offering the best scalar and signal-processing performance of any Cortex-M. That’s on top of the memory protection unit and instruction set brought on by the Armv8-M architecture that optimizes loops and branches.
It also hails the M-Profile Vector Extension, an optional architectural extension comprising 150 new scalar and vector instructions that optimize operations commonly used in machine learning and digital signal processing applications. MVE supports half- and double-precision floating-point operations with vector integer. Put simply, it helps leverage the 9-to-10 stage vector and floating-point pipeline in the Cortex-M85 to process more data per clock cycle, thereby accelerating AI at the edge and digital signal processing applications.
STM32V8: Leveraging innovative embedded non-volatile memory (eNVM) with phase-change memory
The need for a new memory
Using an 18-nm process node posed numerous challenges, including the use of flash memory. Indeed, using a floating-gate memory (such as NAND or NOR) is not cost-efficient at this advanced CMOS node.
Consequently, ST had to find a new memory that was more practical with such technology, while also offering a competitive NVM cell area and reliability similar to flash. That turned out to be our phase-change memory (PCM). In essence, PCM uses a chalcogenide compound that changes state (from amorphous to crystalline and vice versa) in response to a temperature change induced by a current pulse. An amorphous state is highly resistive, while its polycrystalline phase is highly conductive. By applying the proper current, the system can use the phase change to store a “0” or “1”.
The smallest eNVM cell in an MCU
ST is already using phase-change memory in our Stellar family of automotive MCUs. The expertise we gained in this very demanding market translated well to general-purpose microcontrollers, leading us to use PCM on the STM32V8 when we needed to use such a small process node. In fact, our PCM allows us to halve the memory footprint compared to MRAM and RRAM, thus offering the smallest non-volatile memory cells in an MCU. Our PCM supports 10,000 write cycles for code, 100,000 cycles for data, and provides 20 years of data retention. Moreover, since it doesn’t require erase operations like a flash module, it can improve efficiency.
Proven track record
ST’s phase-change memory in the STM32V8 was one of the main reasons SpaceX chose the STM32V8 for their Starlink constellation. Low Earth Orbit environments are particularly challenging. The company thus needed a microcontroller robust enough to withstand high-radiation conditions while offering real-time processing capabilities and benefitting from a large embedded memory resistant to magnetic fields. SpaceX has already announced using the STM32V8 in the Starlink mini laser system for high-speed, low-latency connectivity, with speeds of up to 25 Gbps at 4,000 kilometers. Additionally, SpaceX shared that its teams are also looking at other applications that could benefit from the new ST microcontroller.
STM32V8: serving and safeguarding with graphics and security certifications
While the STM32V8 comes with significant optimizations for machine learning applications, it remains a general-purpose MCU. Hence, our teams ensured that it could serve a wide range of graphical applications by including our Chrom-ART graphical IP and a JPEG Codec. As a result, developers can accelerate 2D copy operations and transparency effects, among other things, to create convincing graphical user interfaces. The STM32V8 also comes with an LCD display controller. If developers use only internal RAM, they can drive a 600 x 480 screen, but can go up to 1280 x 800 with external RAM. We will even provide an STM32V8 Discovery Kit to help teams get started with their TouchGFX project.
Another essential criterion for a general-purpose microcontroller is security. This is why the STM32V8 distinguishes itself from the previous STM32H7 series by being able to target PSA Level 3 and SESIP Level 3 certifications, just like on the STM32H5, STM32N6, STM32U5, and more. Teams can even enjoy the many features in the STM32 Trust framework or hide protect level (HDPL) mechanisms to isolate the boot loader from attacks. Additionally, teams can aim for the latest regional certifications in Europe, the Americas, and more. We want to ensure that every new STM32 series is not only equipped with our cryptocore to accelerate encryption and decryption operations, but that the entire industry objectively recognizes the safeguards we bake into our solutions.
