STM32N6 has been officially launched

[Julian E.]

The STM32N6 has been officially launched yesterday at the STM32 Summit! 

 

 

The STM32N6 is based on the Arm® Cortex®-M55 running at 800 MHz. It is the first STM32 MCU to embed the ST Neural-ART accelerator™, an in-house developed neural processing unit (NPU) engineered for power-efficient edge AI applications.

 

A dedicated computer vision pipeline with a MIPI CSI-2 interface and image signal processing (ISP) ensures compatibility with a wide range of cameras. The STM32N6 also features an H264 hardware encoder and the NeoChrom™ Accelerator for graphics, making it suitable for feature-rich products.

 

4.2 Mbytes of contiguous embedded RAM are also offered, ideal for neural networks or graphics applications, complemented by high-speed external memory interfaces (hexa-SPI, OCTOSPI, FMC).

 

The full AI ecosystem is out as well!

• ST Edge AI Core v2.0: introducing the STM32N6 with Neural Art, and also the Stellar P
• ST Edge AI Developer Cloud supporting the STM32N6 with Neural Art, and also the Stellar P
• X-CUBE-AI v10.0: STM32Cube Expansion Package based on ST Edge AI Core v3.0
• STM32 model zoo v3.0: with many new models and use cases for the STM32N6
• STM32N6 application packages: getting started packages, AI H264 UVC, optimized application such as hand landmark, pose estimation, person detection and power measurement
• STM32 ISP IQTune v 1.01: STM32N6 and STM32MP23 supported hardware in addition to STM32MP25
• Hugging Face organization with ST model cards and application to run the model zoo in Hugging Face
• NVIDIA TAO: image classification notebook for N6
• Updated Wikis and online documentation

 

You can find these tools in the ST Edge AI Suite.

 

Product pages:

• STM32N6 series and lines:
• https://www.st.com/en/microcontrollers-microprocessors/stm32n6-series.html
• https://www.st.com/en/microcontrollers-microprocessors/stm32n6x5.html
• https://www.st.com/en/microcontrollers-microprocessors/stm32n6x7.html
• STM32N6 boards:
• https://www.st.com/en/evaluation-tools/stm32n6570-dk.html
• https://www.st.com/en/evaluation-tools/nucleo-n657x0-q.html
• STM32N6 Cube :
• https://www.st.com/en/embedded-software/stm32cuben6.html

[TDJ]

Great news, especially more MHz and SRAM MB.
For a brief moment I thought that I could use STM32N657X0H3Q for my current project instead of STM32U5G9.
Although FDCAN is upgraded (or, rather STM32U5 FDCAN was downgraded) it looks like there is no MIPI-DSI interface, no DACs, only two 12b ADCs. For my current project it is a showstopper since 1024px (o higher) LCD displays with RGB interface practically do not exist.

[onderdelen]

What a disappointing new STM32 .......

No flash, only RAM, too fast to do anything useful but obviously designed to serve as the core for AI spy-on-people-with-cameras devices. I will never ever consider this device for systems I design.

 

 

 

[TDJ]

@onderdelen I agree, disappointment is the right word. The more I think about it, probably lack of internal flash is the biggest concern. Sure, one can add external flash but what about security? Code would have to be stored encrypted and then decrypted when copying to internal SRAM for execution which would surely eliminate any benefits of extra SRAM size. I am trying to make any sense out of this ST strategy, but I just cannot.
STM32U5G9 with just one MB SRAM extra and full FDCAN implementation (see here) would make me happy.
More MHz, 16b ADC and Cortex-M55 core (in this order) would make me very happy. Am I asking for too much?