Hardware: LoongArch, GNU/Linux Devices, and ESP

posted by Roy Schestowitz on Dec 23, 2025

• LoongArch

  I recently saw the news that LoongArch64 is becoming officially supported in Debian. For those of us who have been around the free software block a few times, the name "Loongson" brings back memories of the Lemote Yeeloong - that small, underpowered netbook that RMS used for years because it was the only laptop that could run with a 100% free BIOS.

• Linux Gizmos ☛ Luxonis OAK 4 CS Edge-Inference Camera with CS-Mount Optics and PoE

  The OAK 4 CS is built around Luxonis’ RVC4 vision compute platform, combining a 6-core ARMv8 CPU, 8 GB of RAM, and 128 GB of onboard storage. The product description states that the platform delivers up to 52 TOPS of AI inference performance using DSP and GPU acceleration for on-device vision workloads.

• Linux Gizmos ☛ AAEON Introduces 3.5-inch SubCompact System with Multi-M.2 and RAID Support

  The GENESYSM-MTH6 is built around Intel Core Ultra processors (Series 1, formerly Meteor Lake), with options ranging from 15 W U-series to 28 W H-series SKUs. Supported processors include the Core Ultra 5 125U/125H and Core Ultra 7 155U/155H, providing up to 16 cores and integrated Intel Arc graphics for compute- and graphics-intensive edge applications.

• CNX Software ☛ ESP-Scope is a web-based oscilloscope built using the ESP-IDF framework and Gemini 3 LLM

  ESP-Scope is an open-source firmware transforming any ESP32 board into an oscilloscope using one of the ADC pins up to 83,333 Hz sample rate (on the ESP32-C6) and visualizing the results over Wi-Fi in a web browser, be it Surveillance Giant Google Chrome, Mozilla Firefox, or others.

• Linux Gizmos ☛ Auvidea X242 Carrier Board Brings Dual 10GbE and PCIe Gen 5 to Jetson T5000

  The X242 is designed as a scalable, industrial-grade carrier board intended for commercial deployment volumes. It is built to host NVIDIA’s Jetson T5000 module, which integrates a Blackwell-architecture GPU with 2,560 CUDA cores and 96 fifth-generation Tensor Cores, delivering up to 2,070 TFLOPS (FP4 sparse) of AI performance.