Skip to main content

AI Overview

Core 3588E runs AI inference on the NPU built into its RK3588 SoC. Mixtile OS includes the Rockchip runtime that runs models on it.

The NPU is programmed with Rockchip's RKNN toolkits. Models and applications made for NVIDIA Jetson, with CUDA, TensorRT or DeepStream, do not run on Core 3588E; a model has to be converted to the RKNN format first.

What runs where​

Rockchip's model conversion toolkit runs on an x86-64 Linux computer. Core 3588E loads the converted model and runs it on the NPU.

On a computerOn Core 3588E
Convert a model, for example from ONNX, to .rknn with RKNN-Toolkit2, for the rk3588 targetLoad the .rknn model with the RKNN runtime, librknnrt, and run inference

Example programs can be built on Core 3588E itself, so no cross-compiler is needed. The YOLOv8 example downloads a model that is already converted, so you can run it without a computer.

On Mixtile OS​

  • The NPU is enabled and librknnrt is installed. You do not need to install an NPU driver or runtime package first. To check the NPU driver:

    Core 3588E
    sudo cat /sys/kernel/debug/rknpu/version

    The output looks like RKNPU driver: v0.9.8.

  • The examples use Rockchip's C/C++ API. Rockchip's Python package for on-board inference, RKNN-Toolkit-Lite2, is provided for Python 3.12 and earlier, which matches the Ubuntu 24.04 and Debian 12 editions of Mixtile OS. Debian 13 has Python 3.13 and Ubuntu 26.04 has Python 3.14, so the package cannot be installed there.

  • Guides written for other RK3588 boards often install vendor packages or enable the NPU with a board-specific tool. Those steps do not apply to Mixtile OS.

Mixtile OS also includes Rockchip multimedia components, such as RGA, MPP, FFmpeg and GStreamer, used by some application workflows.

Keep versions compatible

The RKNN runtime on the module and the RKNN-Toolkit2 version a model was converted with must be compatible. Mixtile OS has librknnrt 2.3.2. If a model you converted does not load, convert it again with RKNN-Toolkit2 2.3.2.

Start with a verified example​

Run YOLOv8 on the Core 3588E NPU​

Build Rockchip's YOLOv8 example with the RKNN C/C++ runtime and detect objects in an image. See Run YOLOv8 on the Core 3588E NPU.