Focused on Rockchip solutions development

RK3576 AI Terminal Mainboard

RK3576 AI Terminal Mainboard

The RK3576 AI Terminal Mainboard is an industrial-grade AI mainboard based on the Rockchip RK3576 8nm processor. Designed for AI facial recognition turnstiles, edge computing boxes, self-service terminals, and industrial HMI systems, it serves as an upgraded next-generation solution over the RK3288.

Its key advantage is the built-in 6 TOPS hardware NPU, which enables the mainboard to perform facial recognition, object detection, and lightweight large AI model inference locally, without relying on an external AI camera.

This product is an AI terminal mainboard based on the Rockchip RK3576 octa-core processor, with a maximum clock speed of 2.2GHz. It integrates a powerful 6 TOPS AI accelerator NPU, supporting hardware acceleration for the latest Transformer-based large AI models and a variety of deep learning frameworks.

The board runs Android 14 by default and also supports Linux and OpenHarmony. It delivers 4K@120Hz ultra-high-definition display output and provides a rich range of expansion interfaces. It is primarily designed for edge computing, AI robots, AI cloud servers, and high-performance edge AIoT devices.

Compared with the RK3288 turnstile mainboard, the RK3576 solution adds high-speed interfaces and enhanced AI camera capabilities:

  1. Display Output: HDMI 2.1 (4K@120Hz), eDP, LVDS, and MIPI-DSI. Supports three independent displays and can directly drive 7–15-inch touchscreen displays.
  2. Camera Interfaces: Multiple MIPI-CSI-2 and USB 3.0 interfaces, supporting simultaneous connection of multiple binocular facial recognition cameras.
  3. Network Connectivity: Dual Gigabit Ethernet ports; Wi-Fi 5 + Bluetooth 5.2; M.2/mini-PCIe interfaces supporting 4G/5G modules.
  4. Industrial Control: Relay output, Wiegand 26/34, RS485, RS232, CAN-FD, multiple GPIO and ADC interfaces, hardware watchdog, and wide-voltage 9–24V power input. Some configurations support PoE power supply.
  5. High-Speed Peripherals: USB 3.0, USB 2.0, and PCIe 2.1, with support for external AI accelerator cards.
Main FunctionSpecification
Board Dimensions146 × 104 mm
CPURK3576, Octa-Core (4 × Cortex-A72 + 4 × Cortex-A53), up to 2.2GHz
GPUARM Mali-G52 MC3
NPU6 TOPS
Operating SystemAndroid 14
Memory / Storage4GB standard (2GB / 8GB optional) / 32GB standard (16GB / 64GB / 128GB optional)
Built-in ROM2KB EEPROM (not included by default; optional)
DP Output1 × DP via Type-C (OTG); supports 1/2/4 lanes at 1.62/2.7/5.4/8.1Gbps; up to 4K×2K@120Hz
LVDS Output1 × LVDS; supports single/dual-channel 6/8-bit; directly drives 50/60Hz LCD panels; up to 1920×1080@60Hz
eDP Output1 × eDP; directly drives eDP LCD panels with various resolutions; tested up to 1920×1080. According to the SoC datasheet, supports up to 4K@60Hz
Audio Input / OutputSpeaker output with left/right channel support; 6W standard, configurable to 10W (resistor modification required); 1 × MIC IN
Headphone Output1 × 3.5mm 4-pole headphone jack
USB Interfaces1 × Type-C OTG; 1 × USB 3.0 HOST; 5 × USB 2.0 HOST
Serial Ports2 × RS232 (both configurable as TTL); 1 × RS485 (configurable as TTL); 4 × TTL UARTs (2 configurable as RS232)
I²C Interface1 × I²C interface for I²C touch panels or other peripherals
Network10/100/1000M adaptive Ethernet; built-in Wi-Fi and Bluetooth; built-in Mini PCIe interface for 4G connectivity (voice calls not supported)
StorageSupports USB flash drives and storage expansion; TF card not supported
RTCLow-power real-time clock
System UpgradeSupports local USB system upgrade
Optional Functions1 × CAN interface; 1 × MIPI camera interface
  1. Android 14 — Ideal for facial recognition turnstiles, self-service terminals, and other smart terminal applications.
  2. Linux — Supports Debian 12, Ubuntu 24.04, and Buildroot.
  3. OpenHarmony and RT-Linux for industrial applications.

Provides a complete BSP and RKNN SDK, enabling customers to perform secondary development and deploy customized applications.

  1. Next-Generation Facial Recognition Turnstile Mainboard: Uses the onboard NPU to perform facial recognition locally, eliminating the need for an external AI camera module. It can simultaneously handle Wiegand, relay control, and display functions, making it an ideal upgrade to legacy RK3288 solutions.
  2. AI Edge Computing Boxes: Suitable for object detection, behavior analysis, and private deployment of local AI models.
  3. Self-Service Terminals, Visitor Management Systems, Vending Machines, and Industrial HMI Systems.
  4. Smart Commercial Displays and Multi-Channel Video Capture Terminals.
  1. Built-in Hardware NPU: Enables local AI inference for facial recognition and other AI algorithms, simplifying the overall system architecture and reducing peripheral costs.
  2. Powerful Performance: Built on an 8nm process with an octa-core 64-bit architecture, delivering a significant performance upgrade over the RK3288. Android 14 also provides a newer and more secure software platform.
  3. Rich Industrial Interfaces: Combines advanced multimedia capabilities with industrial interfaces such as CAN-FD, Wiegand, and relay control, making it ready for turnstile and industrial control applications.
  4. High-Speed Expansion: Supports PCIe high-speed storage, USB 3.0, and multiple camera inputs.
  5. Comprehensive Software Support: Mature SDK ecosystem and stable supply from multiple board manufacturers, providing reliable support for commercial product development and mass production.

Recent Products

Whatsapp
Whatsapp
Email
Email
wechat
wechat
wechat

Get a Quote

Our product experts and technicians will answer your questions within 24 hours.

We use cookies to ensure that we give you the best experience on our website.