製品詳細
Wiki & 学習
レビュー
PDF をダウンロード

特長

The reComputer RK3576 Module is a compact, production-ready compute module designed for seamless integration into custom carrier boards and industrial hardware solutions. Powered by the Rockchip RK3576 SoC, it bridges the gap between mid-range cost and high-end performance, delivering robust Octa-core processing, a dedicated 6 TOPS NPU, and superior 8K multimedia capabilities for next-generation edge intelligence.

Feature 1: High-Performance Compute & Flexible Configurations

  • Advanced Processing Subsystem: Octa-core CPU architecture featuring 4× Cortex-A72 @ 2.2GHz + 4× Cortex-A53 @ 2.0GHz, paired with an ARM Mali-G52 MC3 GPU to handle demanding multi-threaded industrial control and automation tasks.
  • Flexible Hardware Configurations (Optional): Available in multiple memory and storage combinations to perfectly align with your specific edge workload requirements:

Memory: 4GB / 8GB LPDDR5 (Scalable up to 16GB).

Storage: 32GB / 64GB eMMC(Scalable up to 128GB).

 

Feature 2: Next-Gen Edge AI Engine & Production-Ready Software Ecosystem

  • 6 TOPS NPU Acceleration: Integrated hardware AI accelerator supporting INT4/INT8/INT16/FP16/BF16/TF32 mixed quantization operations. This powerful core enables efficient on-device AI inference, eliminating cloud dependencies while ensuring data privacy and low-latency responses.
  • reComputer AI Lab Supported (One-Command Deployment): Streamline your entire development pipeline from PyTorch or TensorFlow to RKNN via Seeed’s optimized toolchain. Skip tedious build environments and achieve one-command or no-code Docker deployments in minutes.

Comprehensive AI Workload Support: Pre-optimized and validated for modern multimodal Edge AI workloads:

  • Computer Vision (CV): Deep integration with YOLOv11, achieving real-time high-frame-rate low-latency inference at up to 77.9 FPS (640×640), perfect for object detection, segmentation, defect detection, and intelligent video analytics.
  • Large Language Models (LLM): Fully capable of running optimized lightweight local language models, including DeepSeek-R1-Distill-Qwen (1.5B/7B) for offline industrial copilots, AI assistants, and knowledge retrieval.
  • Vision-Language Models (VLM): Supports next-gen multimodal models like Qwen2-VL for complex visual reasoning and intelligent tracking.
  • Speech AI: Enables completely offline voice interaction, automated speech recognition (ASR), and Text-to-Speech (TTS) using advanced frameworks like Whisper.

Check the details of the benchmark.

 

Feature 3: Industrial-Grade System-on-Module (SOM) Architecture

  • Optimized for Carrier Board Integration: Tailored specifically for direct physical embedding into proprietary B2B custom carrier boards, moving past mere generic development and validation hardware.
  • 200-Pin High-Density B2B Connectors: All native I/O interfaces are securely routed via standard 200-pin high-density board-to-board (B2B) connectors, ensuring an ultra-compact SOM footprint while maintaining premium signal integrity for industrial deployments.

Feature 4: Extensive Interface Expansion via IO Board

While keeping the module layout highly compact for system integration, the underlying architecture reserves rich native signals, enabling the full reComputer RK3576 Evaluation Kit carrier board to unlock desktop-class and industrial connectivity:

  • Display & Video: HDMI 2.1, MIPI DSI, and USB Type-C (DP Alt Mode) supporting simultaneous 3-display outputs.
  • Camera Input: MIPI CSI (Up to 4-lane).
  • Wireless Connectivity: Dual-band Wi-Fi 6, Bluetooth 5.x, with miniPCIe expansion capabilities for 4G LTE, LoRaWAN, and Wi-Fi HaLow.
  • High-Speed Peripherals: PCIe 2.1 (1-lane), M.2 M-Key (Supports NVMe SSDs and external AI accelerators up to an expanded 26 TOPS).
  • Industrial Networking: 2x Gigabit Ethernet (PoE supported via evaluation carrier board).

The module's pinout architecture is also compatible with the Raspberry Pi CM4 IO Board for baseline development.

Feature 5: Unlock the Official Raspberry Pi Expansion Ecosystem

The reComputer RK3576 Module supports both the Raspberry Pi CM4 IO Board and the Seeed R3576 IO Board. When paired with the Seeed Studio IO Board, it also supports compatible Raspberry Pi peripherals.

  • Vision: Raspberry Pi Camera Module 3
  • Display: Official 7-inch DSI Display
  • Power & Expansion: PoE+ HAT and other compatible accessories

Build flexible solutions for machine vision, interactive displays, and Power over Ethernet applications.

Feature 6: Industrial Reliability & OS Ecosystem

  • 24/7 Continuous Operation: Built from the ground up for high stability, long lifecycles, and uninterrupted runtime under full computational edge workloads in demanding environments.
  • Commercial Grade: 0°C to 70°C operating temperature range (Optimized for robust enterprise systems, indoor kiosks, and standard commercial terminals).
  • Official Armbian Support & Linux Ecosystem & Android 14+: Fully compatible with Linux (Pre-installed with Armbian, supporting Debian and Ubuntu) 、Android 14 and . Provides long-term maintained system images, regular security updates, OTA support, and a stable SDK to drastically reduce time-to-market.

Feature 7: 4K Multimedia & Ultra-HD Processing

  • Advanced Video Processing Unit (VPU): Engineered for heavy video decoding and multi-channel high-resolution vision systems.
  • Video Decoding: Up to 8K @ 30fps (H.265 / H.264 / AV1 / AVS2).
  • Video Encoding: Up to 4K @ 60fps (H.265 / H.264).

仕様

ParameterDescription
Hardware Specification
Product SeriesreComputer RK3576 Module
SoCRockchip RK3576
CPU4 × Cortex-A72 @ up to 2.2 GHz + 4 × Cortex-A53 @ up to 2.0 GHz
GPUArm Mali-G52 MC3
NPUUp to 6 TOPS INT8; supports INT4 / INT8 / INT16 / FP16 / BF16 / TF32 mixed-precision operations
Video DecodingUp to 8K@30fps or 4K@120fps; H.265, VP9, AV1, AVS2, and H.264
Video EncodingUp to 4K@60fps; H.264 and H.265
RAMLPDDR5: 4GB / 8GB (Scalable up to 16GB)
Onboard StorageeMMC: 32GB / 64GB (Scalable up to 128GB)
Module Interface
Board-to-Board Connector2 × 100-pin B2B connectors
Carrier Board CompatibilitySeeed RK3576 IO Board & Raspberry Pi CM4 IO Board
Display InterfacesHDMI/eDP, MIPI DSI, and USB-C DP
Camera Interfaces4-lane + 2-lane MIPI CSI-2
PCIePCIe 2.1 ×1
USBUSB 3.2 Gen 1 and USB 2.0
EthernetGigabit Ethernet
Storage ExpansionSDMMC and PCIe
Peripheral InterfacesGPIO, I²C, SPI, UART, PWM, CAN, and I²S/PCM
Wireless Communication
Wi-FiOnboard Wi-Fi 6
BluetoothBluetooth 5.4
AntennaU.FL / I-PEX connector for an external FPC antenna
Software Ecosystem
Operating SystemsArmbian (pre-installed), Debian, Ubuntu, Android 14+, and balenaOS support
AI ToolchainRKNN Toolkit2 and reComputer AI Lab
Electrical and Environmental
Power Input5V DC
Operating Temperature0–70°C, subject to the carrier board and system thermal design
Heat DissipationExternal cooling recommended
Form Factor55 mm × 40 mm
Others
Production LifetimeUntil January 2036
Warranty1 year
StatementInterfaces require a compatible carrier board*

ハードウェア概要

用途

  • Smart Security & Video Analytics: Multi-channel Network Video Recorders (NVR), intelligent video security boxes, AI-powered crowd analytics, and perimeter intrusion detection.
  • Industrial IoT & Automation: Industrial PCs (IPC), Protocol Gateways, Programmable Logic Controller (PLC) data aggregators, HMI touch panels, and predictive maintenance nodes.
  • Robotics & Physical AI: Autonomous inspection drones/rovers, voice-controlled robots, visual navigation, and AI agents with physical interaction.
  • Smart Retail & Commercial Terminals: AI interactive kiosks, self-service vending terminals, smart digital signage with audience analytics, and POS edge intelligence.
  • Energy & Smart Infrastructure: Smart grid monitoring units, EV charging station controllers, environmental monitoring stations, and intelligent traffic telemetry gateways.
  • Medical & Healthcare: Medical image display assist terminals, health monitoring data aggregators, and laboratory automation control modules.

認証

HSCODE8473309000
USHSCODE8473309100
UPC
EUHSCODE8471707000
COOCHINA

部品リスト

reComputer RK3576 Modulex1

reComputer RK3576 Module: High-Performance Rockchip Compute Module for Industrial Edge AI | 6 TOPS NPU | 200-Pin B2B Connector | Up to 16GB LPDDR5 RAM and 128GB eMMC

SKU 100027907

It’s a high-performance Rockchip compute module for industrial edge AI, powered by an octa-core CPU and 6 TOPS NPU. With a 200-pin B2B connector, rich I/O, flexible LPDDR5/eMMC configurations, and 8K video decoding, it accelerates AI vision, robotics, and intelligent edge devices.

JPY¥21367.00
Shipping and duties calculated at checkout.
  • Free Insurance Included On Us
  • Flexible Payments with PayPal Pay Later
  • Avoid Import Fees: Ship Tax-Free from Local Warehouses
  • 1 on 1 Support | 12-Month Warranty | 30-Day Return Policy