Embodied AI Glossary中文

Rockchip RK3588

瑞芯微 RK3588Advanced

An 8nm, eight-core ARM chip from Rockchip with a 6 TOPS NPU, a common low-cost robot compute platform.

The RK3588 is a flagship SoC (a chip integrating a CPU, GPU, AI accelerator, and more) from Fuzhou-based Rockchip, built on an 8nm process, with a CPU combining 4 Cortex-A76 and 4 Cortex-A55 cores, a Mali-G610 GPU, a built-in 6 TOPS NPU (neural-network accelerator), and support for 8K video encode/decode. It's used in a huge number of dev boards and industrial mainboards, and in robots it commonly serves as a low-cost, low-power onboard compute platform or host computer, running ROS 2, camera processing, and lightweight model inference; a model has to be converted with RKNN-Toolkit before it can run on the NPU. Compared with NVIDIA's Jetson Orin, it's cheaper and more power-efficient, but its AI compute and CUDA ecosystem lag far behind, so it can't handle a large VLA model.

ExampleA YOLO detection model is converted to RKNN format with RKNN-Toolkit and deployed to the NPU on an RK3588 dev board, giving a robot car real-time object recognition.

Also called
RK3588, RK3588S
Related
Onboard Compute Platform · Rockchip RKNN-Toolkit · Neural Processing Unit (NPU) · NVIDIA Jetson · On-Device / Edge Deployment · TOPS (Tera Operations Per Second)
Sources
Rockchip RK3588 - Rockchips.net
CNX Software: Rockchip RK3588 specifications revealed
As of
2026-09

See it in the full glossary →