Sipeed Brings Kendryte's K210 Edge Neural Network Accelerator to PCIe M.2, USB Type-C with MAIX Nano

Kendryte's popular K210 RISC-V system-on-chip (SoC) with neural processing unit (NPU) has found a new home in the Sipeed MAIX Nano M1n, an ultra-compact deep learning acceleration board based on the PCI Express M.2 form factor — and shipping with a USB Type-C adapter to boot.

Designed to accelerate edge neural networking workloads, the Kendryte K210 includes a dual-core 400MHz 64-bit RISC-V processor with hardware floating-point unit alongside a custom convolutional neural network processing unit (NPU, styled by Kendryte as the Kendryte Processing Unit or KPU) coprocessor. The accelerator has its own memory, with 6MB of static RAM given over to the dual-core CPU and a further 2MB for the NPU coprocessor.

While it's becoming increasingly common to see single-board computers shipping with the K210, CNX Software has highlighted a new product out of Sipeed which brings the chip to any system with PCI Express or USB Type-C connectivity. The Sipeed M1n takes the K210 and pairs it with with 128Mb of SPI flash plus a 24-pin edge connector for a compatible camera — everything you need for VGA-resolution computer vision tasks at up to 30 frames per second.

For those looking to do more with the part, the MAIX Nano M1n includes through-hole headers for 16 GPIO pins plus LCD display connectivity while also converting the M.2 form factor board into one accessible over USB Type-C connectivity. As with other K210-based devices, the M1n is compatible with common deep-learning frameworks and can be programmed using the Arduino IDE, Platform.io, or MicroPython.

The Sipeed MAIX Nano M1n is scheduled to begin shipping late this week, priced at just $9.90 from Seeed Studio.