Home / Internet Of Things / Kneron unveils new ultra low-power AI processors

Kneron unveils new ultra low-power AI processors

The American society of complex generation and in-depth studying Kneron unveiled a spread of low-power synthetic intelligence (AI) processors.

Consistent with the corporate, the gadgets each and every eat lower than five mW of power, in contrast to maximum equivalent processors that require a couple of watts to function.

The Kneron NPU IP Sequence is composed of the ultra-low chronic model KDP 300, the KDP Usual Model 500, and the high-performance model KDP 700.

The brand new processors are designed to run AI packages for smartphones, and for smartphones, safety and different IoT gadgets.

With them, customers are in a position to take advantage of deep studying networks, reminiscent of ResNet and YOLO at the gadgets, says the corporate.

Learn extra: MIT Develops Low-Power High-Speed ​​Chips for IoT Security [19659007] According to Kneron, they offer 'comprehensive hardware solutions for advanced AI, including the most advanced hardware solutions. IP hardware, compiler and model compression ".

Processors work with models of neural networks such as Resnet-18, Resnet-34, Vgg16, GoogleNet and Lenet.

Sophisticated Technology

New chips use so-called decomposition technology to "divide a large-scale convolutional compute block into a smaller number of blocks for parallel computing".

Kneron explained: "With reconfigurable convolutional acceleration technology, small block calculation results will be integrated to achieve better overall computing performance.

" Advanced Model Compression Technology from Kneron, the size of non-optimized models can now be reduced by a few dozen times. The multi-level caching technique reduces the use of CPU resources and further improves overall operational efficiency. "

The NPU IP-KDP 300 enables facial recognition in 3D, while the NPU IP-KDP 500 can be used for analysis.And finally, the high-performance NPU IP-KDP 700 chip can handle an IT AI more advanced.

Albert Liu, founder and CEO of the company, said, "To perform AI computing on peripherals, the biggest challenge is to achieve performance while maintaining low power consumption."

"Kneron's NPU IP Series provides high performance at very low power, which is a breakthrough in IA Edge."

Read more: ARM Launches Scalable Chips for machine learning IoT

Learn extra: Why you might soon have a neural network on your smartphone

Web of Trade says

The peripheral atmosphere is turning into extra vital for the propagation and the IoT luck as a result of increasingly important selections must be made, and purposes carried out, at velocity and on the level of beginning or within reach. As intelligence and computational chronic are an increasing number of driven into networks and peripherals, speeds want to build up and effort intake must lower, as do prices.

Kneron's innovation on this area is a part of this sectoral pattern. with the improvement of low chronic safety chips and an increasing number of optimized for AI packages and system studying.

About zahir

Check Also

actility abeeway cisco and tracktio show iot geolocation in the port of barcelona - Actility, Abeeway, Cisco and Tracktio show IoT geolocation in the Port of Barcelona

Actility, Abeeway, Cisco and Tracktio show IoT geolocation in the Port of Barcelona

This is the record at the deployment and trying out of a location and monitoring …

Leave a Reply

Your email address will not be published. Required fields are marked *