Built around Amlogic's 6nm A311Y3 Edge AI SoC, the CM311Y3 System-on-Module delivers a significant leap in performance-per-watt over previous-generation A311-series platforms. Its octa-core CPU adopts a big.LITTLE architecture with 2x Arm Cortex-A78 performance cores and 6x Arm Cortex-A55 efficiency cores, complemented by a shared 2MB L3 cache. Two integrated RISC-V coprocessors handle system control, always-on power management, and sensor hub duties - offloading housekeeping tasks from the main CPU cores to maximize application performance.

At the heart of the CM311Y3 lies an 8 TOPS Amlogic Deep Learning Accelerator (ADLA) supporting INT4, INT8, INT16, FP8, FP16, and BF16 precision. Compatible with TensorFlow, TensorFlow Lite, ONNX, PyTorch, Darknet, MxNet, and Caffe, the NPU enables low-latency inference for object detection, facial recognition, and behavioral analysis directly at the edge. This dedicated AI acceleration frees the CPU and GPU from tensor operations, delivering responsive AI experiences without cloud dependency.

The CM311Y3 is engineered for demanding multimedia and connectivity workloads. The ARM Mali-G625 MC1 GPU supports OpenGL ES 3.2, Vulkan 1.4, and OpenCL 3.0, while the AVE-10 video engine handles 4K@120 fps decoding (AV1, H.265, VP9, AVS3) and 4K@60 fps H.265/H.264 encoding. The module integrates HDMI RX/TX for video capture and display, dual Ethernet for reliable networking, PCIe 3.0 for high-speed peripheral expansion, and MIPI CSI/DSI for camera and display connectivity - providing a complete interface suite for embedded system design.

With up to 16GB LPDDR5X memory and 256GB eMMC 5.1 storage, the CM311Y3 is ideally suited for 4K video surveillance systems, AI edge computing devices, intelligent interactive terminals, enterprise thin clients, and autonomous robots - delivering proven, low-latency AI performance in a compact, production-ready module.

 PCB Dimension

  Pin Definition

If you order the SOM or any question, please feel free to contact us at market@armdesigner.com