RK3588 SBC: 6 TOPS NPU Rockchip Single Board Computer for Edge AI & Industrial Vision

Сайт RK3588 SBC (single board computer) is Rockchip’s flagship embedded computing platform, purpose-built for high-performance edge AI, industrial vision and multimedia applications. Built on the 8nm Rockchip RK3588 SoC, this board combines an octa-core Arm CPU, a dedicated 6 TOPS NPU and an 8K-capable video engine on a single compact board.

For industrial solution providers and embedded developers, the RK3588 single board computer solves a critical problem: it delivers server-class AI inference and video processing at the edge, without relying on cloud connectivity or expensive x86 industrial PCs. This makes it a preferred hardware choice for smart factories, autonomous robots, multi-camera inspection systems and intelligent gateways.

In this guide, we break down the RK3588 SBC specifications, real-world application scenarios, development ecosystem and how it compares to lower-tier Rockchip boards like the RK3568.

RK3588 SBC Core Specifications

The RK3588 SoC at the heart of this SBC is Rockchip’s highest-performance general-purpose processor, designed for AIoT and edge computing workloads. Below are the verified hardware specifications based on Rockchip’s official datasheet:

Source: Rockchip RK3588 official brief datasheet (V1.0, 2025); Forlinx OK3588-C product specification; Radxa NX5 module documentation.

What Makes the RK3588 SBC Different

Three hardware features set the RK3588 SBC apart from mid-tier embedded boards:

1. Dedicated 6 TOPS NPU: The triple-core NPU handles AI inference locally, eliminating cloud latency and privacy concerns. It supports mixed-precision operations (INT4 to TF32), allowing developers to balance accuracy and speed for models like YOLO, MobileNet and ResNet.
2. 8K Video Engine: The integrated VPU decodes 8K@60fps and encodes 8K@30fps, making it suitable for high-resolution surveillance NVRs, multi-camera vision systems and 4K/8K digital signage.
3. Multi-Display & Multi-Camera: The RK3588 can drive 3–4 independent displays and process 4–8 camera streams simultaneously — a critical capability for industrial HMIs, robotics and smart kiosks.

Key Application Scenarios for RK3588 SBC

The RK3588 single board computer is not a general-purpose hobbyist board. It is engineered for commercial and industrial deployment. Below are the highest-value use cases:

1. Edge AI Inference & Computer Vision

The 6 TOPS NPU makes the RK3588 SBC ideal for on-device AI tasks such as:

• Object detection and classification (YOLOv5/v8, SSD)
• Face recognition and access control
• Defect inspection on production lines
• Pose estimation for robotics
• License plate recognition (LPR)

Unlike cloud-based solutions, edge inference on the RK3588 delivers sub-100ms response times and operates without internet connectivity — essential for factory floors and remote installations.

2. Machine Vision & Automated Optical Inspection (AOI)

With its 32MP ISP, multi-camera MIPI inputs and 8K video decoding, the RK3588 SBC powers vision inspection systems that:

• Capture high-resolution images from multiple industrial cameras
• Run real-time defect detection using the NPU
• Output results to industrial HMIs or PLCs via RS485/CAN
• Store inspection logs locally on NVMe SSD

This makes it a cost-effective alternative to expensive x86-based industrial vision PCs.

3. Industrial Control & Smart Gateway

The RK3588 SBC’s rich industrial I/O (RS485, CAN, GPIO) and dual Gigabit Ethernet make it suitable as:

• Industrial IoT gateway: Collecting data from PLCs, sensors and Modbus devices, then uploading to cloud platforms via MQTT or OPC UA
• Edge controller: Running real-time control logic for conveyor systems, AGVs and smart manufacturing equipment
• Protocol converter: Translating between industrial protocols (Modbus RTU/TCP, CANopen, EtherCAT)

4. Multi-Camera NVR & Video Surveillance

The 8K VPU and multi-camera ISP allow the RK3588 SBC to function as a high-channel-count NVR:

• Decode and record 16+ 1080p camera streams simultaneously
• Run AI-powered motion detection and object tracking on the NPU
• Support 4K/8K output for monitoring walls
• Store footage on NVMe SSD or network storage

For mid-to-large surveillance deployments, the RK3588 outperforms lower-tier SoCs like the RK3568, which typically handles only 4–8 camera streams.

5. Robotics & Autonomous Systems

The RK3588 SBC is increasingly used as the main compute module in:

• Роботы AGV/AMR: Running SLAM navigation, obstacle detection and motion planning
• Service robots: Powering voice interaction, face recognition and autonomous navigation
• Industrial robot arms: Handling vision-guided picking and quality inspection

The combination of NPU (AI perception), GPU (3D rendering for HMI) and industrial I/O (motor control) makes it a single-board solution for robotics.

6. Digital Signage & Smart Kiosk

With 8K output, multi-display support and strong media decoding, the RK3588 SBC powers:

• 4K/8K digital signage players
• Multi-screen video walls
• Interactive kiosks with touch displays
• Smart retail terminals with AI-powered customer analytics

RK3588 SBC vs RK3568 SBC: Which Should You Choose?

A common question among embedded developers is whether to use the RK3588 or the more cost-effective RK3568. The answer depends entirely on your workload:

Система принятия решений

Choose RK3568 SBC if:

• Your project requires basic HMI, data logging or simple IoT gateway functionality
• AI workloads are light (e.g., basic sensor anomaly detection)
• Cost and power consumption are primary concerns
• You only need 1–2 cameras and a single display

Choose RK3588 SBC if:

• You need real-time AI inference (object detection, face recognition, defect inspection)
• Your system processes multiple camera streams (4+)
• You require 4K/8K video output or multi-display support
• You are building robotics, AOI systems or high-end industrial HMIs
• You want performance headroom for future feature upgrades

Related reading: RK3568 vs RK3588: Which Rockchip Development Board Fits Your Industrial Project?

Development Ecosystem & Software Support

The RK3588 SBC benefits from Rockchip’s mature SDK ecosystem and active community support:

Операционные системы

• Ubuntu 20.04/22.04: Full desktop and server images, ideal for AI development and general computing
• Debian 11/12: Lightweight, stable for headless industrial deployments
• Android 12/13: For smart displays, kiosks and consumer-facing devices
• Buildroot / Yocto: For minimal, customized firmware in mass-produced products

AI Development Tools

• RKNN Toolkit2: Rockchip’s official model conversion and deployment toolchain. Converts TensorFlow, PyTorch, ONNX and Caffe models to RKNN format for NPU acceleration.
• Support for common frameworks: TensorFlow Lite, PyTorch Mobile, ONNX Runtime (with RKNN delegate)
• Pre-trained model zoo: YOLOv5/v8, MobileNet, ResNet, SSD, FaceNet and more, optimized for RK3588 NPU

Firmware & Customization

• Complete kernel source code (Linux 5.10+) and U-Boot bootloader
• Device tree customization for custom carrier boards
• OTA (over-the-air) update support for field deployment
• Industrial-grade firmware with long-term support (LTS) from Rockchip and board vendors

Practical guide: How to Flash Rockchip Firmware: A Step-by-Step Tutorial for RK3568/RK3588 Boards

Hardware Design Considerations for Industrial Deployment

When integrating the RK3588 SBC into a commercial product, consider the following:

Thermal Management

The RK3588 consumes 10–20W under full load (CPU + NPU + VPU active). Industrial designs require:

• Active cooling (heatsink + fan) for sustained AI workloads
• Thermal throttling configuration in the device tree
• Wide-temperature-range components for industrial environments (-20°C to 70°C)

Источник питания

• Recommended 12V/3A or 5V/4A DC input with robust power regulation
• Consider UPS or battery backup for critical industrial applications
• Power sequencing for peripherals (NVMe, cameras, displays)

Storage Reliability

• Use industrial-grade eMMC or NVMe SSD for write-intensive applications
• Enable filesystem journaling and read-only root filesystem options for reliability
• Consider RAID or network backup for NVR and data-logging applications

ЧАСТО ЗАДАВАЕМЫЕ ВОПРОСЫ

Q: What does SBC stand for in RK3588 SBC?

SBC stands for Одноплатный компьютер. It is a complete computer built on a single circuit board, with microprocessor, memory, input/output and other features required for a functional computer. The RK3588 SBC integrates all these components using the Rockchip RK3588 SoC.

Q: How much AI performance does the RK3588 NPU deliver?

The RK3588’s triple-core NPU delivers 6 TOPS (trillion operations per second) of AI computing power. In practical terms, it can run YOLOv5s object detection at 60+ FPS on 1080p video, or MobileNetV2 image classification at 200+ FPS — all on-device, without cloud connectivity.

Q: Can the RK3588 SBC run Windows?

No. The RK3588 is an Arm-based processor and does not support Windows natively. It runs Linux distributions (Ubuntu, Debian), Android and embedded RTOS. For Windows-based industrial applications, an x86 SBC would be required.

Q: Is the RK3588 SBC suitable for mass production?

Yes. Many hardware vendors offer industrial-grade RK3588 SBCs and system-on-modules (SOMs) with stable long-term supply. The RK3588 SoC itself is in active production and widely used in commercial products ranging from smart displays to industrial edge devices. For mass production, consider a SOM + custom carrier board approach for cost optimization.

Q: What is the difference between RK3588 and RK3588S?

The RK3588S is a cost-optimized variant of the RK3588 with reduced peripheral interfaces (fewer PCIe lanes, display outputs and camera inputs), but retains the same CPU, GPU and NPU performance. The full RK3588 is preferred for industrial applications requiring maximum expandability, while the RK3588S is suitable for cost-sensitive consumer or light industrial products.

Q: How do I deploy AI models on the RK3588 SBC?

Use Rockchip’s RKNN Toolkit2 to convert your trained model (TensorFlow, PyTorch, ONNX, Caffe) to RKNN format. The toolkit supports quantization (INT8/INT16) to optimize performance and memory usage. Once converted, deploy the RKNN model using the RKNN Runtime API on the RK3588, which leverages the NPU for hardware-accelerated inference.

Заключение

Сайт RK3588 SBC represents the sweet spot for high-performance edge computing in industrial and AIoT applications. With its octa-core CPU, 6 TOPS NPU, 8K video engine and rich industrial I/O, it delivers capabilities that were previously only available in expensive x86 industrial PCs — at a fraction of the cost and power consumption.

Whether you are building an edge AI vision system, a multi-camera NVR, an industrial IoT gateway, an autonomous robot or a high-end digital signage player, the RK3588 single board computer provides the performance, expandability and software ecosystem needed to take your product from prototype to mass production.

For projects with lighter workloads and tighter budgets, the RK3568 SBC remains a solid choice. But when AI performance, multi-camera processing or 4K/8K multimedia is on your requirements list, the RK3588 SBC is the Rockchip platform to choose.

Ready to evaluate the RK3588 SBC for your project? Contact our embedded hardware team for technical consultation and customized solution design.