Our NPU is an AI computing engine developed based on years of processor architecture research and development capabilities. It is designed specifically for edge devices. The biggest advantage of this NPU is that it can provide efficient deep learning reasoning capabilities.

In order to obtain faster response time and higher computing performance, this NPU adopts the same frequency design as the main CPU, which can perform complex neural network reasoning tasks at a faster speed. At the same time, this NPU adopts low-power technology, which enables it to have excellent computing power even in resource-constrained environments. In addition, low-bit quantization technology, that is, by converting the entire neural network model to a lower bit width, greatly reduces the storage requirements of the model and reduces computing overhead, thereby reducing energy consumption and optimizing energy efficiency without affecting reasoning performance.

Combined with the AI operator development platform, it supports mainstream network frameworks such as Caffe, TensorFlow, and PyTorch. The platform has built-in multiple optimization algorithms and a rich model zoo to facilitate user development and use.