In order to optimize your browsing experience we uses cookies. You agree to the usage of cookies when you browsing this site.
For more information regarding cookies and the processing of your personal data please read our Privacy Policy.
SUPPORT
Listen to Your Voices and Let Us Know Your Needs
Back
Gowin NPU
IP Introduction
Gowin NPU IP, embedded neural network processing unit, is used to accelerate to compute machine learning algorithms parallelly, including AHB bridge unit, accelerator unit, PSRAM controller unit, and SPI Flash controller unit.
AHB bridge unit, which is bridged with MCU through AHB bus, is used to realize the cooperation between NPU and MCU.
Accelerator unit, which is calling machine learning algorithm, is used to accelerate to compute AI model parallelly. The machine learning algorithms supported by accelerator unit include deep convolution algorithm (DepthwiseConv2D), two-dimensional convolution algorithm (Conv2D), maximum pooling algorithm (MaxPool2D) and average pooling algorithm (AveragePool2D).
PSRAM controller unit, is auxiliary accelerator unit to store intermediate temporary calculation data. When the accelerator unit calculates the AI model of each layer, PSRAM inputs the data of the previous layer of the model, and outputs the calculation result data of the current layer to the next layer after the calculation of the current layer is completed.
The SPI Flash controller unit, is auxiliary accelerator unit to store the Weight data and Bias data of the AI model. When the accelerator unit computes the AI model of each layer, the weight data and bias data of the current layer of the model are read from the SPI Flash and used to calculate the machine learning algorithm of the current layer of the model.