D-Robotics RDK X3 4GB Development Kit Cortex-A53 X3M BPU (Brain Processing Unit) 5Tops

TheD-Robotics RDK X3is an ultra-compact AI development board powered by theSunrise X3M SoCwith adual-core Bernoulli-architecture BPU (Brain Processing Unit)delivering5 TOPSof edge inference performance. Built around aquad-core ARM Cortex-A53 @ 1.5 GHzprocessor with4GB LPDDR4memory, the RDK X3 provides a Raspberry Pi-compatible (85 56 mm) platform optimized for lightweight AI workloads, robotic perception, and cost-sensitive embedded applications.

Despite its compact size and sub-$100 price point, the RDK X3 punches well above its weight class. The integrated BPU accelerates neural network inference forYOLO object detection, image classification, human pose estimation, and VIO (Visual-Inertial Odometry) all supported out of the box through theNodeHubapplication center, which provides over 100 open-source algorithms and ready-to-deploy applications. WithH.265 hardware decoding up to 4K@60fps, dual MIPI CSI camera interfaces, HDMI output, Gigabit Ethernet, and a 40-pin GPIO header compatible with Raspberry Pi expansion boards, the RDK X3 covers the essential connectivity requirements for vision-based robotics and edge AI projects.

The RDK X3 is available in three configurations Single Board,Basic Kit(with case, power, and accessories), andDev Kit(adding an 8MP IMX219 camera and CSI adapter board) to accommodate every stage of your development workflow, from bare-board integration to full vision system prototyping.

Features

• Sunrise X3M SoC with 5 TOPS BPU Dual-core Bernoulli-architecture Brain Processing Unit delivers efficient edge AI inference for object detection, classification, segmentation, and pose estimation at ultra-low power consumption
• Quad-Core Cortex-A53 @ 1.5 GHz Efficient application processor for running Linux, robot control loops, and concurrent AI inference tasks with minimal thermal footprint
• 4GB LPDDR4 Memory Sufficient for running lightweight YOLO models, stereo vision, and standard robotics middleware simultaneously (2GB variant also available for cost-optimized deployments)
• H.265 (HEVC) Hardware Decode 4K@60fps High-efficiency video decoding for multi-stream camera processing, enabling real-time video analytics with minimal CPU overhead
• Dual 2-Lane MIPI CSI Camera Interfaces Connect up to two MIPI cameras for stereo depth estimation, dual-view perception, or simultaneous RGB + depth sensing via the optional Dev Kit
• HDMI 1.4 Output 1080P@60fps Display output for HMI visualization, debug monitoring, and real-time inference result preview
• 1 USB 3.0 + 2 USB 2.0 Host USB 3.0 for high-bandwidth peripherals (cameras, LiDAR); USB 2.0 ports for mice, keyboards, and general accessories
• Gigabit Ethernet (RJ45) 1000Mbps wired networking for low-latency data transfer, ROS 2 communication, and remote development access
• Wi-Fi (2.4/5 GHz, 802.11 b/g/n) + Bluetooth 4.2 Wireless connectivity for untethered robot operation, OTA updates, and wireless peripheral pairing
• 40-Pin GPIO Header (Raspberry Pi Compatible) Drop-in compatible with a wide ecosystem of Raspberry Pi HATs, pHATs, and expansion boards, supporting PCM/I2S signal protocols
• Ultra-Lightweight Only 29g Ideal for weight-sensitive applications including drones, small-form-factor robots, and portable edge devices
• Ultra-Wide Temperature Range: -25C ~ 95C Suitable for extreme environments including outdoor robotics, automotive applications, and industrial installations
• NodeHub 100+ Open-Source Algorithms One-click deployment of pre-built applications including traffic light detection, VIO, humanoid robot soccer control, toy car line tracking, and deep learning autonomous driving
• Ubuntu 20.04 / 22.04 Support Standard Linux environment with familiar toolchain, enabling seamless integration with ROS 2, OpenCV, Python, and C++ development workflows
• CE / MIC / SRRC Certified Fully compliant with international electromagnetic compatibility and radio regulatory standards

Specifications

Multimedia

Connectivity

USB Ports

Expansion & Debug

Power & Physical

Variant Comparison

RDK X3 vs RDK X5 Quick Comparison

Reference Documents:

• D-Robotics Developer Community Documentation & SDK Downloads
• RDK Model Zoo Pre-trained AI Models (GitHub)
• MIPI Camera Driver for RDK (GitHub)
• RDK X3 Product Page (Waveshare)

Application

• Entry-Level Robotic Perception Deploy lightweight YOLO models on the 5 TOPS BPU for real-time object detection, human tracking, and obstacle avoidance in mobile robots and rovers
• Vision-Based Autonomous Driving Leverage H.265 4K@60fps decode and dual MIPI cameras for lane detection, traffic sign recognition, and deep learning-based line-following applications
• Visual-Inertial Odometry (VIO) Run VIO algorithms out of the box via NodeHub for precise position estimation in GPS-denied environments, ideal for drone and indoor robot navigation
• Raspberry Pi Replacement / Upgrade Drop-in compatible 40-pin GPIO header allows reuse of existing Raspberry Pi HATs and pHATs while adding powerful AI inference capabilities
• Edge AI Prototyping Ultra-light 29g form factor and low 15W power envelope make the RDK X3 ideal for UAV payloads, portable devices, and space-constrained embedded systems
• Education & STEM NodeHub’s 100+ pre-built applications (soccer robot control, toy car tracking, traffic light detection) provide ready-made curricula for AI and robotics education
• Smart Camera & Video Analytics H.265 hardware decoding combined with NPU inference enables multi-stream video analysis for security, monitoring, and quality inspection
• ROS 2 Development Ubuntu 20.04/22.04 support enables seamless ROS 2 integration for sensor fusion, navigation, and multi-robot coordination

Packing List

Single Board

Basic Kit

Dev Kit

FAQ

Q1: What is the difference between the Single Board, Basic Kit, and Dev Kit?
TheSingle Boardincludes only the bare RDK X3 board (4GB) suitable for integration into existing systems or for developers who already have compatible accessories. TheBasic Kitadds a protective case (RDK-Case-A), UART debug module, 4-pin cable, 64GB TF card with pre-loaded OS, card reader, and a 5V/3A power adapter with multi-region plug adapters. TheDev Kitincludes everything in the Basic Kit plus an8MP IMX219-77 camera module (79.3 FOV, with FFC cable)and anX3 Pi CSI adapter board everything needed for vision and AI application development out of the box.

Q2: How does the RDK X3 compare to the RDK X5?
The RDK X3 is theentry-level, cost-optimizedsibling of the X5. Key differences: the X3 has aquad-core Cortex-A53 @ 1.5 GHzvs the X5’s octa-core Cortex-A55 @ 1.8 GHz;5 TOPSBPU vs 10 TOPS;Wi-Fi 4 / BT 4.2vs Wi-Fi 6 / BT 5.4; no PoE or CAN FD; and a single USB 3.0 port vs four. However, the X3 offers awider operating temperature (-25C ~ 95C), a significantly lower board weight (29g), and lower power consumption (15W vs 25W). Choose the X3 for lightweight, cost-sensitive applications; choose the X5 for maximum AI performance and connectivity.

Q3: Can I use Raspberry Pi HATs with the RDK X3?
Yes. The RDK X3 features a40-pin GPIO header with Raspberry Pi-compatible pinoutand supports PCM/I2S signal protocols. Many Raspberry Pi expansion boards, sensors, and displays can be connected directly. However, electrical compatibility should be verified for specific HATs due to potential voltage or signal differences between platforms.

Q4: What AI models and algorithms are available?
TheNodeHubapplication center provides100+ open-source algorithms and applicationsdeployable with one click, including traffic light detection, visual-inertial odometry (VIO), humanoid robot soccer control, toy car line tracking, and deep learning autonomous driving. Additionally, theRDK Model Zooon GitHub offers pre-trained YOLO and other models optimized for the X3 BPU.

Q5: Does the RDK X3 support ROS 2?
Yes. The RDK X3 runsUbuntu 20.04 / 22.04, which provides native support for ROS 2 Humble and Foxy. Combined with the available MIPI camera drivers (hobot_mipi_cam), you can build complete ROS 2 perception and navigation pipelines.

Q6: What cameras work with the RDK X3?
The Dev Kit includes theIMX219-77 camera module (8MP, 79.3 FOV)with an X3 Pi CSI adapter board. The board provides2 2-lane MIPI CSI interfaces. Third-party MIPI cameras compatible with the hobot_mipi_cam driver may also work. For stereo vision, two cameras can be connected simultaneously one to each MIPI CSI port.

Q7: How do I power the RDK X3?
The RDK X3 requires5V / 3A (15W)power via the USB Type-C port. The Basic Kit and Dev Kit include a compatible 5V/3A power adapter with US/EU/UK plug adapters. When using the Single Board, ensure your power supply can deliver at least 5V/3A. Note: unlike the RDK X5, the X3 doesnotsupport Power-over-Ethernet (PoE).

Q8: Can the RDK X3 decode 4K video?
Yes. The Sunrise X3M SoC includeshardware H.265 (HEVC) decoding up to 4K@60fps (Main Profile L5.1). This enables efficient multi-camera video processing and real-time analytics with minimal CPU load, making it suitable for video surveillance, streaming, and vision-based AI applications.

Q9: Is the 2GB variant available?
Yes. The RDK X3 is available in both2GBand4GB LPDDR4variants. The 2GB version is suitable for lighter workloads such as basic YOLO inference, simple robot control, and headless Linux applications. The 4GB version is recommended for multi-camera setups, larger AI models, and memory-intensive perception stacks.

Q10: What is the warranty policy for the RDK X3?
Please refer to the manufacturer’s warranty policy as specified on the product packaging or official documentation. For warranty claims and technical support, contact the point of purchase or visit theD-Robotics Developer Community.