En
Heterogeneous Compute for Efficient Edge Inference
The CM5 quad-core A76 handles the OS and pre-processing, while the Hailo-8 runs neural network inference independently. This offloads the CPU and increases the number of vision tasks a single unit can handle.
BCM2712 at 2.4GHz with up to 16GB LPDDR4 provides solid general-purpose performance.
Hailo-8 delivers 13 TOPS for mainstream vision models; object detection and classification run entirely on the edge.
Full support for AI framework deployment under Raspberry Pi OS with a mature model conversion workflow.
Fanless Dual-Path Cooling: Full Load Without Throttling
A single aluminium heatsink covers all heat sources. The fanless design reduces failure points and is suitable for long-term operation in cabinets and production lines.
Thermal pads on the CPU, Hailo NPU, power supply and Wi-Fi module prevent hot spots.
At 60°C with both CPU and Hailo-8 fully loaded, clock speeds remain stable with no performance loss.
Dual Ethernet: Separate Data and Control Traffic
Two independent Ethernet ports — 2.5GbE and 1GbE — isolate data traffic from control traffic for better stability.
The 2.5GbE port handles high-bandwidth workloads such as camera streams and model updates.
The 1GbE port connects PLCs, gateways and remote maintenance on a separate, more secure network.
USB 3.0 5Gbps and HDMI 2.0 4K@60Hz cover peripherals and local display.
Wide-Input Power and Rugged Design for Harsh Sites
The power input and mechanical design are optimised for industrial environments, reducing the impact of external conditions.
9–28VDC wide-input range with TVS transient protection; works with standard 24V factory supplies.
Metal enclosure with DIN rail mount; fanless with no moving parts for high vibration resistance.
Onboard RTC for time keeping and a crypto chip for license management.
Raspberry Pi Open-Source Ecosystem: AI Frameworks and Tools Ready to Use
Reliable system: Raspberry Pi OS has long-term maintenance and ongoing security updates, supporting cyber resilience requirements such as CRA.
Tools out of the box: GPIO libraries, OpenCV, Node-RED and more install and run directly, shortening development time.
AI integration: mainstream deep learning frameworks such as YOLO can be deployed directly and work with Hailo-8 for edge vision inference.
Active community: the global Raspberry Pi developer community is mature and active; most development problems already have proven solutions available online.
Typical Applications
AI edge computing terminals
Machine vision
Industrial control
Data acquisition gateways
Thin clients
