what is the use of npu

NPU is a kind of processor specially applied to network application data packets. It adopts the architecture of “data-driven parallel computing” and can be used to process massive multimedia data such as video and image.

The nup on the mobile phone can be used as a neural network processing unit. It is a professional chip constructed by simulating biological nerves. It is born for deep learning and enables mobile phones to obtain stronger computing and processing capabilities in the field of artificial intelligence.

NPU is also a kind of integrated circuit, but different from the single function of special-purpose integrated circuit, the network processing is more complex and more flexible. Generally, software or firmware can be specially programmed according to the characteristics of network operation to realize the special purpose of the network. Many different functions are implemented on one chip to speed up the efficiency of processing network packets, and are used in many different network devices and products.

NPU advancement process

On June 20, 2016, the State Key Laboratory of Digital Multimedia Chip Technology of Zhongxing Microelectronics announced in Beijing that it has successfully developed China’s first embedded neural network processor (NPU) chip, becoming the world’s first embedded device with deep learning artificial intelligence. Video capture and compression coding system-level chip, and named “Xingguang Intelligent No. 1”. This deep learning-based chip is used in face recognition, with a maximum accuracy rate of 98%, exceeding the recognition rate of the human eye. The chip was mass-produced on March 6 this year, and the current shipment is more than 100,000 pieces.

Zhang Yundong, executive director of the laboratory and chief technology officer of Zhongxingwei, said in an interview that chips equipped with neural network processors are used in surveillance cameras, and the cameras have been upgraded from “eyes” to “eyes with brains”, which is a global phenomenon. first. The State Key Laboratory of “Digital Multimedia Chip Technology” was established in 2010, relying on Beijing Zhongxing Microelectronics Co., Ltd. and approved by the Ministry of Science and Technology. According to data, Zhongxing Microelectronics Co., Ltd. was established in 1999 by direct investment from the former Ministry of Information Industry of the People’s Republic of China. It is a “national team” among companies specializing in chip technology. core” situation. The landing of artificial intelligence “Xingguang Smart One” is an embedded NPU. The neural network processor NPU (Neural Processing Unit) is not yet well known, but it is a popular technology in the chip field. Compared with the CPU processor in the von Neumann architecture, it adopts a subversive new architecture of “data-driven parallel computing”. If the von Neumann architecture processes data as a single lane, then “data-driven parallel computing” is 128 multi-lane parallel, which can process 128 data at the same time, which is conducive to processing massive multimedia data such as videos and images.

In the industry, the computing performance per unit power consumption, that is, the performance-per-watt ratio, is used to measure the pros and cons of the processor architecture. According to Zhang Yundong, executive director of the laboratory and chief technology officer of Zhongxingwei, the performance and power consumption ratio of “Xingguang Smart One” is “at least two or three orders of magnitude higher” than the traditional Von Neumann architecture, that is, hundreds of times.

High power consumption is criticized by many top artificial intelligence technologies. IBM’s “Deep Blue” in the 20th century and Google’s AlphaGo in 2016 are supported by huge data calculations. The former uses supercomputers and the latter uses server clusters, which cannot be separated from the computer room of constant temperature and humidity. AlphaGo costs $3,000 to play a game of chess. Zhang Yundong called them “a scientific experiment”, and there is still a long way to go before the technology is implemented and put into application.

This highlights the miniaturization, low power consumption and low cost advantages of the embedded NPU, and accelerates the application of artificial intelligence technology. For example, drones have high requirements on the weight and power consumption of cameras, otherwise it will affect take-off and endurance. The “Starlight Smart No. 1” is only the size of an ordinary postage stamp and weighs only a few tens of grams. Its birth has enabled many small devices such as surveillance cameras to have the possibility of artificial intelligence, and has taken artificial intelligence from the mysterious computer room to life applications. step.
Responsible editor: YYX

Leave a Reply

Your email address will not be published.