Introduction

The reComputer RK3576 is powered by the high-performance RK3576 processor (4x Cortex-A72 @2.2GHz + 4x Cortex-A53 @2.0GHz) and an ARM Mali-G52 MC3 GPU. This device offers 4GB, 8GB, or 16GB of LPDDR5 RAM and features an NPU with 6 TOPS of computing power, supporting INT4/INT8/INT16/FP16 operations. In terms of multimedia, it supports video encoding up to 4K@60fps and video decoding up to 8K@30fps. Regarding interface requirements, it is equipped with two Gigabit Ethernet ports (one of which supports PoE), along with onboard Wi-Fi 6 and Bluetooth 5.4. The USB interfaces include 1x USB 3.0, 3x USB 2.0, and 1x Type-C port supporting OTG and DP output. It features three display outputs: Display, Type-C (DP 1.4), and a 4-lane MIPI DSI, and comes with an M.2 Key M 2280 slot (PCIe 2.1 x1) for an NVMe SSD or AI accelerator. Thanks to its dual Gigabit Ethernet, onboard wireless communication, and multiple display interfaces, this device is highly suitable for deployment in smart retail terminals, lightweight edge gateways, and building automation scenarios.

Get One Now

Features

• Robust Hardware Foundation & Scalable Computing Power
  • Core Processor: Powered by the Rockchip RK3576, featuring an Octa-core architecture (4x Cortex-A72@2.2GHz + 4x Cortex-A53@2.0GHz) alongside an Arm Mali-GPU.
  • High-Speed & Large Capacity Memory: Supports up to 16GB LPDDR5, ensuring exceptional memory throughput for on-device Large Language Models (LLMs) and demanding AI workloads.
  • Built-in & Expandable NPU: Delivers 6 TOPS of native on-device AI performance. Features a PCIe 2.1 (M.2 M Key) slot for hardware acceleration expansion, scaling up to 26 TOPS.
• Comprehensive AI Model Support & Multimodal Capabilities
  • Covers Three Major AI Scenarios: Purpose-built to handle Perception AI, Generative AI, and Physical AI applications.
  • Seamless On-Device Execution of Mainstream Models:
    • Visual Interaction: Achieves 77.9FPS running YOLOv11 (640x640), and fully supports vision-language models (VLMs) like Qwen2.5-VL 3B, as well as MobileNet, RetinaFace, and CLIP.
    • On-Device LLMs: Natively supports the deployment of DeepSeek-R1-7b.
    • Real-time Voice Interaction: Features low-latency Speech-to-Text (Whisper-base 2.0s) and Text-to-Speech real-time processing capabilities.
  • Robust Underlying Frameworks: Broad compatibility with mainstream frameworks including ONNX, PyTorch, TensorFlow, and Caffe, supported by the mature RKNN-Toolkit2 toolchain.
• Versatile Network Connectivity & Peripheral Expansion
  • Comprehensive Network Communication: Equipped with 2x Gigabit Ethernet ports, Wi-Fi 6, and Bluetooth 5.4. Additionally supports PoE, 4G, LoRaWAN, Wi-Fi Halow, and SIM card expansion for diverse IoT deployments.
  • Exceptional Hardware Expansibility: Integrates a rich array of I/O interfaces, including MIPI-CSI, MIPI DSI, DP1.4, USB 3.0, PCIe 2.1, CAN FD, UART, PWM, I2C, and SPI.
• Flagship Multimedia & Display Output
  • Versatile Display Interfaces: Provides four types of display connectivity options, including HDMI 2.1, DP 1.4, MIPI-DSI, and USB Type-C, offering flexible interface selections for independent or duplicated multi-screen configurations.
  • Hardware-Accelerated Codec: Features an H.265 / H.264 / AV1 / AVS2 multivideo decoder supporting up to 8K@30fps, paired with an H.264 / H.265 multivideo encoder supporting up to 4K@60fps for efficient media streaming.
• Frictionless Software Ecosystem & One-Click Deployment
  • Out-of-the-Box OS: Comes pre-installed with Armbian and offers Multiple OS Support.
  • reComputer AI Lab Ecosystem: Engineered for rapid deployment with minimal complexity. Utilize simple Docker commands to instantly pull and run containers (e.g., launching deepseek-r1-distill-qwen:1.5b in seconds). Explore the full range of supported pre-configured environments on the https://sensecraft.seeed.cc/ai-lab/models .
  • Extensive Developer Resources: Grants access to over 100 ready-to-use AI models, 100+ open-source project libraries, and comprehensive, step-by-step tutorials.

Specifications