Forlinx has launched the FCU3101, a fanless edge AI computing box built around the Rockchip RV1126B or industrial-grade RV1126BJ SoC, aimed at IIoT and smart security use cases like video surveillance, license plate recognition, intrusion detection, and material counting. It offers up to 4GB DDR4, up to 64GB eMMC, dual Ethernet, Wi-Fi 4, Bluetooth 4.2, optional 4G LTE, RS485/RS232, digital I/O, relay outputs, and a wide 9-36V DC input with industrial temperature tolerance from -40°C to +85°C. The quad-core Cortex-A53 SoC includes an NPU delivering up to 3 TOPS INT8, supporting TensorFlow, ONNX, PyTorch, and Caffe models. It runs Linux with ready-to-use AI vision algorithms, though the exact kernel version isn't specified (a related Forlinx SoM shipped Linux 6.1.141). No pricing is disclosed; interested buyers must contact Forlinx directly, with a 5-day sample lead time and 6-week bulk lead time, and CE/FCC/RoHS certifications expected by Q4 2026.
Questions this post answers
What are the specs of the Forlinx FCU3101 edge AI box based on Rockchip RV1126B?
The Forlinx FCU3101 is a fanless industrial edge AI gateway using the Rockchip RV1126B or industrial-grade RV1126BJ quad-core Cortex-A53 SoC with an NPU rated up to 3 TOPS INT8. It offers up to 4GB DDR4, up to 64GB eMMC, dual Ethernet, Wi-Fi 4, Bluetooth 4.2, optional 4G LTE, RS485/RS232, relay outputs, and a 9-36V DC input, operating from -40°C to +85°C in its industrial configuration. Track new Rockchip-based edge AI hardware releases like this one on daily.dev.
What AI frameworks does the Rockchip RV1126B NPU support?
The Rockchip RV1126B's NPU, rated up to 3 TOPS at INT8, supports INT4, INT8, INT16, and FP16 model precisions and works with TensorFlow, ONNX, PyTorch, and Caffe frameworks. This makes it suitable for deploying pre-trained or custom vision models for tasks such as personnel detection, license plate recognition, and defect inspection on embedded edge devices. Compare NPU capabilities across edge AI SoCs like this on daily.dev when picking hardware.