Longqing Shixin adopts a high-performance edge-side universal neural network accelerator for large model inference on general-purpose edge-side architectures. Based on the self-developed EE-DNA (Edge/End Diverse Neural Networks Accelerator) IP computing architecture, it is designed for AIPC products and is committed to providing efficient large model inference capabilities at the edge.
-
A high-performance, configurable software-defined AI processor for multiple categories of applications and various types of models at the edge side.
-
A single processor architecture integrates NPU and DSP processing functions, reducing data access bandwidth requirements.
-
Supports code/graphical programming modes similar to CPU programming, enabling quick configuration and ease of use.
-
Supports mainstream AI architectures including visual AI, audio AI, generative AI, and digital signal processing linear algebra computations.
