Rockchip RK3576 AI Dongle
RK3576
Product Description
MODEL: EMB-RK76
Ultra-Compact. Power-Efficient. AI-Ready. The EMB-RK76 is a thumb-size embedded edge AI computer powered by the Rockchip RK3576 AIoT SoC. Housed in an industrial-grade fanless metal case, this 0.078-litre device delivers 6 TOPS of AI performance for enhanced computer vision and real-time inference at the edge, making it ideal for space-constrained deployments demanding 24/7 reliable operation.
6 TOPS AI Performance: The integrated 3-core NPU delivers 6 TOPS for computer vision, object detection, and real-time AI analytics. Supports INT4/INT8/INT16/BF16/TF32 quantization with Rockchip's RKNN SDK to convert models from TensorFlow, PyTorch, ONNX, and other popular frameworks.
● Rockchip RK3576 AIoT SoC with Quad Cortex-A72 and Quad Cortex-A53 ARM cores
● ARM Mali-G52 MC3 GPU with 4K@120fps video decoding (H.265, H.264, VP9, AV1, AVS2)
● Integrated 6 TOPS NPU (3-core architecture) supporting INT4/INT8/INT16/BF16/TF32 quantization
● Onboard 4GB / 8GB / 16GB 32-bit LPDDR5 system memory
● Onboard 64GB / 128GB / 256GB / 512GB eMMC 5.1 OS storage
● Full-HD display output via USB 3.0 Type-C port
● 4K display output via internal HDMI header
● 1x Gigabit Ethernet RJ45 port for reliable wired networking
● USB 3.0 Type-C host port for high-speed peripheral connectivity
● Industrial-grade fanless metal case with no moving parts
● Ultra-compact 0.078 Litre form factor: 66.0 x 55.8 x 21.3 mm
● Wide operating temperature: -20°C to +55°C (to be confirmed)
● DC 12V Type-C power input
● Pre-installed Debian 12 (kernel 6.1.118); Ubuntu community support
● Rockchip RKNN SDK and RKLLM SDK software toolchain
● 10-year product supply life-cycle commitment
Product Specifications
APPLICATIONS
AI Vision Analytics & People Counting
Multi-sensor fusion with CCTV, LiDAR, IR thermal, and ToF depth cameras for real-time people counting, occupancy monitoring, and pedestrian flow analysis in retail, transit, and smart buildings.
Video-Based Fire Detection (VBFD)
AI-powered visual fire and smoke detection for manufacturing facilities, warehouses, outdoor storage, wildfire surveillance, construction sites, data centers, and transportation infrastructure. Faster detection with fewer false alarms than traditional sensors.
Digital Signage & Kiosk
Compact player for retail displays, menu boards, and information kiosks. Supports both standalone local-network content management and cloud-based CMS deployment with Gigabit Ethernet connectivity.
Industrial Automation & SCADA
Factory floor HMI, PLC communication, sensor data aggregation, and machine vision quality inspection in space-constrained industrial environments requiring fanless, reliable 24/7 operation.
Smart Surveillance & Security
Real-time object detection, pose estimation, and face recognition at the edge. Integrates with existing CCTV infrastructure and VMS servers for proactive security monitoring with on-device AI inference.
Edge AI Gateway & Sensor Hub
Lightweight edge analytics hub aggregating data from camera, LiDAR, thermal, and depth sensors. Protocol translation and cloud pre-processing for IoT deployments in smart buildings and Industry 4.0.
| Processor | Rockchip RK3576 AIoT SoC Quad Cortex-A72 + Quad Cortex-A53 ARM cores |
| System Memory | Onboard 4GB / 8GB / 16GB 32-bit LPDDR5 |
| Graphics Engine | ARM Mali-G52 MC3 H.265, H.264, VP9, AV1, AVS2 decoding up to 4K@120fps H.264/H.265 encoding up to 4K@60fps |
| Integrated NPU | 6 TOPS (3-core architecture) Support INT4 / INT8 / INT16 / BF16 / TF32 quantization |
| Mass Storage | Onboard 64GB / 128GB / 256GB / 512GB eMMC 5.1 |
| Video Output | Full-HD via USB 3.0 Type-C port (VP2, up to 1920x1080 @ 150MHz) 4K via internal HDMI header (VP0, up to 4096x2160 @ 1200MHz) |
| USB Ports | 1 x USB 3.0 Type-C host port |
| LAN | 1 x Gigabit Ethernet RJ45 port |
| Internal I/O | Internal HDMI header (VP0, 4K capable) Boot ROM switch Recovery switch Status LED indicator |
| Cooling System | Fanless, industrial-grade metal chassis, no moving parts |
| Dimensions | 66.0 (W) x 55.8 (D) x 21.3 (H) mm 0.078 Litre |
| Operating Temperature | -20°C to +55°C (to be confirmed) |
| Power Input | DC 12V via USB Type-C |
| Supported OS | Debian 12 (kernel 6.1.118) — pre-installed Ubuntu (community support) |
| Software Stack | Rockchip RKNN SDK — AI model conversion and deployment Rockchip RKLLM SDK — Large Language Model support Compatible with TensorFlow, PyTorch, ONNX, Caffe, Darknet, and more |
| Supply Longevity | 10-year product supply life-cycle |
| AI Performance | 6 TOPS (integrated NPU) |
ORDERING PART NUMBER (SKU)
| SKU | BOM | Description | Power Adapter |
|---|---|---|---|
| EMB-RK76-A0 | 250-AB600-000EB | Embedded PC RK3576, LPDDR5 8GB, eMMC 128GB, USB-C 3.0, GbE |
Not included |
| EMB-RK76-B0 | 250-AB600-001EB | Embedded PC RK3576, LPDDR5 4GB, eMMC 64GB, USB-C 3.0, GbE |
Not included |
Downloads
| Product Brochure |  |
| Product Images |  |