RK1820/RK1828

  • RK3588 Soc(Host) : As the core of the system, responsible for task scheduling, resource allocation, and overall control.

  • RK1820/RK1828 Coprocessor(Device) : As an AI computing acceleration unit, we focus on high-performance neural network inference tasks.

  • PCIe : Realize low latency and high bandwidth data exchange between the main control and co processors.

rknn-smi

rknn-smi (System Management Interface) use as RK1820/RK1828 equipment information collection, function configuration, and log management functions.

  • Software Version: rknn-smi -v

  • Hardware Version: rknn-smi info -l

  • Status: rknn-smi info -w

  • Performance: rknn-smi set -t work_mode -s 2

RKNN3

RK1820/RK1828 SDK dictionary:

rknn/
├── rknn3-model-zoo
├── rknn3-runtime
├── rknn3-toolkit
└── rknn-gstreamer-plugins

Model Zoo

Provide deployment examples of classic models on the RK1820/RK1828 platform.

Runtime

The RKNN3 C API is the C language interface for the RKNN3 Runtime.
Developers use C/C++ to develop application programs and deploy model inference through RKNN3 C API.

Toolkit

RKNN3 Toolkit is a development kit that provides users with model transformation, inference, and performance evaluation on the PC platform.
RKNN3 Toolkit different from RKNN-Toolkit and RKNN-Toolkit2.