ST has just launched its most powerful STM32 microcontroller so far, with the STM32V8 family, the first equipped with a Cortex-M85 core (clocked at up to 800 MHz) and manufactured with an 18nm FD-SOI process.
It’s a non-pin compatible update to the STM32H7 family that delivers up to 5,072 CoreMarks, greatly improves Edge AI performance thanks to Arm Helium and MVE, integrates up to 4MB eNVM (Embedded NVM), operates up to 140°C, and adds PCM (phase-change memory) radiation immunity. Two main parts have been launched in different configurations and packages: STM32V863 and STM32V873.
STM32V863/873 specifications:
- Core – 32-bit Arm Cortex-M85 CPU @ up to 800MHz with Arm Helium, Arm MVE, TrustZone…; up to 5,072 CoreMarks
- Multimedia accelerators – Chrom-ART 2D GPU, and JPEG hardware accelerator
- Memory/Storage
- 1.5 MB system SRAM with ECC (partial)
- 8 KB backup RAM
- 192 KB zero-wait state TCM
- Up to 512 KB TCM with ECC
- Up to 4 MB embedded NVM; 2x banks with 2MB capacity and ECC
- 2x 128 KB data NVM
- 256 KB system NVM
- External – Hexa-SPI, Octo SPI, 2x SD/SDIO/MMC, Fast Flexible memory controller (FMC) for DRAM, NOR, NAND
- Display Interfaces
- LCD TFT controller up to 600×480 (2x 16-bit) in internal RAM, up to 1280×800 (2x 16-bit) in external RAM
- Chrom-GRC controller
- HDMI CEC
- Camera internal – 16-bit parallel camera I/F
- Audio
- 2x SAI, 4x I2S (multiplexed with SPI), S/PDIF Rx
- MDF (Multi-Function Digital Filter) with 6 filters) + ADF (Audio Digital Filter)
- Networking – Gigabit Ethernet with TSN
- USB
- 1x USB HD, 1x USB FS with PHYs
- UCPD (USB-C PD) controller
- Other Peripherals
- 3x I2C + 2x I3C
- 5x USART, 5x UART, 1x LPUART
- 3x FD-CAN
- 6x SPI, four I2S-capable
- Analog
- 3x 12-bit ADC, 1x 12-bit DAC
- Die temperature sensor
- GPIO pins
- Timers/Control
- 3x motor control timers
- HR timer, 10x 16-bit timers
- 5x 16-bit low-power timers
- 4x 32-bit GP timers
- 2x watchdog timers, gtx timer
- FMAC + cordic
- RTC
- Debugging – Serial wire debug (SWD), JTAG, Embedded Trace Macrocell (ETM) – TBC
- System/Misc
- GPDMA/HPDMA
- SMPS, LDO (for RAM)
- Security
- Unique immutable ID
- Secure storage (HUK)
- STM32V873 only
- Secure boot/Upgrade (STiRoT)
- Crypto SAES/AES
- Secure debug
- OTF XSPI Enc./Dec.
- Anti-tampering
- Target SESIP3 & PSA L3 certifications
- Supply Voltage – No data
- Packages
- 68-pin VQFN (8 x 8 mm)
- 100-pin LQFP/TFBGA (14 x 14/8 x 8mm)
- 144-pin LQFP (20 x 20mm)
- 176-pin LQFP (24 x 24mm)
- 176+25-pin UFBGA (10 x 10mm)
- 273-pin TFBGA (14 x 14mm)
- Temperature Ranges – -40°C to 140°C
Twenty-four different STM32V8 parts are available with 2 MB or 4 MB eNVM, different security options, and a choice of packages from 68-pin VQFN up to 273-pin TFBGA.
The STM32 Cortex-M85 microcontroller is supported by the ST edge AI ecosystem, including the NanoEdge AI Studio IDE, STM32 Cube.AI, ST Edge AI developer cloud, and various application examples. The STM32V8 is compatible with popular AI frameworks, formats, and tools such as TensorFlow Lite, Keras, ONNX, PyTorch, and MATLAB. The company also plans to offer the STM32V8-DK development kit and the more cost-effective Nucleo-V8 board for evaluation, but those two are still shown as “Coming soon”.
Up to now, Renesas was the only silicon vendor manufacturing Arm Cortex-M85 microcontrollers with nine different RA8 families to date, and ST Microelectronics’ STM32V8 will soon add another option.
The STM32V8 already has one customer: SpaceX. The company leverages the radiation-hardened Cortex-M85 microcontroller in a mini laser found in Starlink satellites encountered in Low Earth Orbit (LEO). Some selected partners have early-stage access to the STM32V8, OEM availability is scheduled for Q1 2026, and broader availability will follow soon after. Check out the “campaign” page and press release for additional details.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress. We also use affiliate links in articles to earn commissions if you make a purchase after clicking on those links.
“process from GlobalFoundries” seems to be a bit misleading?
I think the facility in Grenoble is a joint venture of the two, and they seem to also collaborate with Samsung for this process.
Do you know more details?
I’ve removed that part. Globalfoundries showed up somewhere when I wrote about it, but now I can’t find it anymore in the PR, PDF, or landing page.
FD-SOI sounds suspiciously like an IBM development, one of them AMD licensed and spun it off to GloFo, so GloFo sounds more than credible also a joint venture sounds like a sensible thing to do for ST.
Finally a high performance MCU available in a low feature leaded package, I just hope those variants will also have the same maximum frequency. I’ve been on the hunt for such an MCU for quite a while, but all current ones (at least Cortex-M7 and 800MHz) are BGAs with quite a lot of pins and features, quite difficult to have them on small boards like the Arduino Nano / Raspberry Pi Pico.
What is the use case for such an MCU? Just curious, I guess I’m lacking a bit of imagination 😊
In my case, I need it to run model predictive control (real time optimization) for a rather compact self balancing robot, where most MCUs are not powerful enough but an SBC would be overkill. I am aware that something like a Teensy is pretty capable, but I don’t have the problem fully defined so I would rather have some head room in processing power, without the complexity or scale of large BGA packages (I haven’t found many small yet capable dev boards so I started considering making a simple one myself).
If it wasn’t obvious by now, yes, this is an academia project, which does not compare to an industry need, but I am sure those exist too.
Cool thanks for sharing your project. I guess hard realtime combined with highly demanding calculations is the sweet spot for such MCUs.
STM when are you going to get with the times and finally offer RISC-V options. I keep checking back in your catalog for RISC-V chips but still nada, so have to look at elsewhere at competitors.
Well besides the price and the nerd factor what does a riscv do better than a cortex M?
I also would love such an st riscv but to be honest I’m not sure if it brings something to me other than coolness 😁