STM32MP25 2nd gen STM32

The first of this STM32MP2 series is the STM32MP25, available with single or dual 1.5GHz 64bit Arm Cortex-A35 cores (up to 6,000 DMIPS), running alongside a 400MHz Cortex-M33 embedded core for real-time processing. External 32bit DDR4 and LPDDR4 memories are supported.

“A dedicated NPU [neural processing unit] adds up to 1.35Top/s of computing muscle optimised for edge-AI acceleration in applications such as machine vision and predictive maintenance,” according to ST.

Comms include Gigabit TSN (time-sensitive networking) with a two-port Gigabit Ethernet TSN switch, PCIe, USB 3.0 and CAN-FD peripherals.


Data concentrators and gateways are potential applications. “For example, the MPU can acquire video from a 5Mpixel sensor at 30frame/s, perform analytics with the Edge AI accelerator, and send relevant video, encoding with a hardware encoder,” said ST.

A video-capable 1080p 3D GPU (1.35Top/s) is included, that supports Vulkan real-time graphics for Android applications. The 1080p encoder-decoder has connections including LVDS, 4-lane MIPI DSI, and MIPI CSI-2 for cameras including raw-Bayer image sensors.

“Security features ensuring SESIP Level 3 certification include Arm’s TrustZone architecture and resource-isolation framework,” said ST. “This is complemented with secure key storage, secure boot, a unique device ID in one-time programmable memory, hardware cryptographic engines, and on-the-fly DDR encryption-decryption.”

Operation is over -40 to +125°C, packaging includes a 0.8mm-pitch TFBGA chip-scale package aimed at four-layer PCBs, and STM32MP2 MPUs get ST’s 10 year longevity commitment.

STM32MP25 samples and evaluation boards are available “to select OEM customers”, said ST. Volume production of chips and boards is scheduled to begin in the first half of 2024.

The STM32MPU ecosystem includes an OpenSTLinux distribution with an AI framework (X-Linux-AI). For the included Cortex-M33 core, there are STM32Cube development tools whose firmware will run bare-metal or under an RTOS.

The STM32MP25 product page can be found here