AIO-3576JD4 Low-Power Large-Model Mainboard

Quick Start: WIKI
Tips

Introduction

AIO-3576JD4 powered by the Rockchip RK3576 SOM, an octa-core 64-bit AIOT processor, EC-R3576PC FD features an advanced lithography process to deliver high performance while maintaining low power consumption. It is equipped with an ARM Mali G52 MC3 GPU and a 6 TOPS NPU, supporting the private deployment of large-scale models under the Transformer architecture. With support for 4K@120fps decoding/4K@60fps encoding, the computer boasts a powerful display capability with 4K resolution at a high frame rate of 120 fps. Its industrial-grade metal enclosure and fanless design enable passive cooling. With an external watchdog, it provides industrial-grade stability, making it an excellent choice for AI applications requiring local deployment.

Features

6 TOPS, 4K@120FPS, Rockchip RK3576, UFS2.0
supports LLMs and traditional CNN neural networks
260 PIN SODIMM SOM

Shipping List

SOM: Only Core-3576JD4 SOM
Dev kit: Core-3576JD4 SOM + Carrier Board; 1x PSU

Applications

Large-Model Computer, AI Edge kit

Specification

Basic Specifications	SOC	Rockchip RK3576
	CPU	Octa-core 64-bit processor (4×A72 + 4×A53), up to 2.2GHz
	GPU	G52 MC3 @ 1GHz, supporting OpenGL ES 1.1/2.0/3.2, OpenCL 2.0, Vulkan 1.1 Built-in high-performance 2D acceleration hardware
	NPU	6 TOPS NPU, supporting INT4/8/16/FP16/BF16/TF32 mixed operations
	ISP	The integrated 16-megapixel ISP supports low-light noise reduction, an RGB-IR sensor, and up to 120dB of HDR. The AI-ISP enhances image quality with reduced noise.
	VPU	Decoding: 4K@120fps: H.265/HEVC, VP9, AVS2, AV1; 4K@60fps: H.264/AVC Encoding: 4K@60fps: H.265/HEVC, and H.264/AVC
	RAM	LPDDR4/LPDDR4x (4GB/8GB optional)
	Storage	eMMC (16GB/32GB/64GB/128GB/256GB optional), UFS 2.0 (optional)
	Storage Expansion	1 * M.2 (SATA3.0/ PCIe NVMe SSD expansion, supporting 2242/2260/2280)
	Power	DC 12V (5.5mm * 2.1mm, 12V~24V wide input voltage)
	OS	Android14, Linux OS, and Buildroot
	Software Support	・ The private deployment of ultra-large-scale parameter models under the Transformer architecture, such as Gemma-2B, LlaMa2-7B, ChatGLM3-6B, and Qwen1.5-1.8B ・ Traditional network architectures such as CNN, RNN, and LSTM; a variety of deep learning frameworks including TensorFlow, PyTorch, MXNet, PaddlePaddle, ONNX and Darknet ・ Custom operator development ・ Docker container management technology
	Size	122.89mm * 85.04mm * 22.7mm
	Weight	≈120g
	Environment	Operating temperature: -20℃~60℃ Storage humidity: 10%～90%RH (non-condensing)
Interfaces	Network	Ethernet: 2 * RJ45 (1000Mbps) WiFi: WiFi/BT module expansion available via M.2 E-KEY (2230) interface, supporting 2.4GHz/5GHz dual-band WiFi 6 (802.11a/b/g/n/ac/ax) and BT5.2 4G: 4G LTE expansion available via Mini PCIe (shared with 5G module) 5G: 5G expansion available via M.2 (shared with 4G module)
	Video Input	2 * MIPI CSI DPHY (1 * 4 lanes or 2 * 2 lanes) 1 * MIPI CSI D/C PHY (MIPI DPHY (14 lanes or 2 2 lanes) or MIPI CPHY (3 lanes) )
	Video Output	1 * HDMI2.1(4K@120fps)
	Audio Output	1 * 3.5mm Audio jack, supporting MIC recording, CTIA standard
	Watchdog	External watchdog
	USB	2 * USB3.0、1 * USB2.0
	Other Interface	1 * Type-C (USB 2.0/DEBUG), 1 * FAN (4Pin-1.25mm), 1 * SIM card 1 * dual-row pin header (210-20PIN-2.0mm): USB 2.0, SPI, 2I2C, Line in, Line out, GPIO 1 * Phoenix connector (24Pin, 3.5mm pitch): 1 RS485, 1 * RS232, 1 * CAN 2.0