1 / 2

Kneron NPU IP

Kneron NPU IP Series are neural network processors that have been designed for edge devices. These processors provide high computing performance with low power consumption and are small in size. Kneron NPU IP Series can be applied to smart homes, smart surveillance, smartphones, and wearable devices that have high requirement for low power and space. The entire product consumes under 0.5W and can even drop below 5mW for specific applications.

High energy efficiency

All series reach higher than 1.5 TOPS/W.

Support mainstream deep learning frameworks

Caffe, Keras, TensorFlow, and ONNX.

Low power consumption

Under 0.5W and can be less than 5 mW for specific applications.

An integrated AI solution

Include hardware IP, compiler, and model compression.

Deep compression technology

Compresses not only models but also data and coefficients during computing to reduce memory use. 

Filter decomposition and convolution acceleration

Divide a large-scale convolutional computing block into a number of smaller ones to compute parallelly, and integrate and accelerate the computing results from the small blocks.

CNN model support optimization

Supports diverse CNN models, including Vgg16ResnetGoogleNetYOLOTiny YOLOLenetMobileNet, and Densenet with model specific performance optimization.

Interleaving computation architecture

Enable parallel convolution computing and pooling to improve overall performance. The convolution layer can support both 8 and 16 bits fixed points concurrently.

Adaptive data structure

According to different computing demand, dynamically adjust the data structure to improve MAC efficiency.

Dynamic storage resource configuration

Allows more efficient resource allocation between shared memory and operational memory. Increases storage resource utilization without affecting performance.