What Is NPU in Mobile Chips? Finally Understand After Reading

In early 21st-century sci-fi movies, AI often appeared as robots, showcasing almost “omnipotent” artificial intelligence, using imaginative creativity to depict a technology-filled future world.However, back in the real world, while AI has not yet achieved the superintelligence seen in blockbuster films, it has already begun to enter our lives through devices like smartphones and smart speakers. AI is no longer just a distant and unfamiliar concept.

Currently, AI has become a standard feature in the smartphone market, but whether the AI in smartphones is intelligent enough depends on how powerful the NPU in the smartphone chip is.So, what exactly is NPU? Recently, Huawei’s Kirin officially released a series of comics titled “Understanding Chips Made Easy” to provide a detailed explanation. Let’s take a look!

NPU:The Core Carrier of Mobile AI

Everyone knows that a smartphone cannot operate normally without a SoC chip, which is only the size of a fingernail yet contains everything needed. The various integrated modules work together to support the phone’s functionality, such as the CPU responsible for smooth application switching and the GPU supporting fast game graphics loading, while the NPU (Neural Processing Unit) is specifically responsible for AI computation and application.

In the smartphone SoC, the NPU plays the smartest role and directly affects the strength of the phone’s AI capabilities.In 2017, Huawei launched its self-developed architecture NPU, which, compared to traditional scalar and vector operation modes, uses a 3D Cube for acceleration in matrix operations. Therefore, the amount of data calculated per unit time is larger, and the AI computing power per unit power consumption is stronger, achieving an order-of-magnitude improvement over traditional CPUs and GPUs for better energy efficiency.

Understanding NPU, but how exactly is AI used?

In 2017, AlphaGo defeated Go champion Ke Jie, recognized by the tech industry as a milestone in AI technology. At the same time, the internet buzzed with doomsday theories about AI, and voices against the development of AI technology were incessant. However, technology is a double-edged sword; as long as AI is applied to more scenarios that meet everyday life needs, it can bring us more convenience.

Based on its deep accumulation of edge AI technology, Huawei, in collaboration with developers and partners, has rolled out many AI applications. For example, in 2018, the Kirin 980 achieved AI pose recognition, accurately identifying human skeletal contours even during rapid motion, providing more foundational capabilities for AI application development.

In 2019, the Kirin 990 series achieved AI portrait real-time segmentation capability, accurately separating the subject from the background and replacing the background in real time. Compared to other smartphones’ background replacement functions, the Kirin 990 series can accurately identify each individual and remove them from the frame, providing new gameplay and possibilities for more video-related AI applications.

In addition to the AI capabilities at the chip level, the Huawei Mate 30 series, equipped with the Kirin 990 series, has implemented a series of AI applications. Among them, AI can follow the face in real time without the need for gravity sensing, allowing the phone interface to rotate in real-time with the direction of the face, so you don’t have to repeatedly lock and unlock the screen while watching videos.

AI air control allows for non-touch interactions like scrolling and screenshotting. With AI air control, you can scroll through Douyin while making dumplings during the New Year, and you can browse the web while wearing gloves in winter. This feature is extremely practical.

Behind the AI applications of the Mate 30 series is the support of the Kirin 990 series’ AI capabilities. It is reported that the Kirin 990 series adopts Huawei’s self-developed Da Vinci architecture NPU, innovatively designed with NPU dual large cores and NPU micro cores. The NPU large cores handle high-computation scenarios, such as AI pose recognition and real-time background replacement, while lightweight AI applications like AI follow and AI air control are processed by the NPU micro cores, occupying fewer resources and further enhancing the performance and energy efficiency of the NPU.

Moreover, while bringing AI smart applications to consumers, Huawei is also vigorously promoting an open AI ecosystem. Based on Huawei’s Da Vinci architecture NPU and HiAI 3.0, the Kirin platform provides developers with a more powerful and complete toolchain and stronger edge computing power. As of December 2019, HiAI 3.0 had nearly 100 million daily active users and over 600 billion monthly calls.

Now entering 2020, AI will also ride the fast train of 5G, achieving a leap in computing power and capabilities, truly opening up an intelligent world of interconnected devices. It is believed that under the leadership of technology giants like Huawei, smartphone AI will become smarter and more practical, bringing us more impressive AI features.Through this comic, do you understand NPU now?

What Is NPU in Mobile Chips? Finally Understand After Reading

Disclaimer:This article is original by the author, and the content reflects the author’s personal views. The Electronics Enthusiast Network reproduces it only to convey a different perspective and does not represent the Electronics Enthusiast Network’s endorsement or support of this view. If there are any objections, please feel free to contact the Electronics Enthusiast Network.

Read More Exciting Articles

Xiaomi’s New Year First Battle! All Mi 10 Models Support Wi-Fi 6! Wi-Fi 6 Enters the Popularization Year
Overall Solution for R&D Testing of IoT Products
Keysight U5855A Infrared Thermal Imager Helps You Develop Sharp Eyes
Frequent Acquisitions of WiFi Chips, 2020 Will Spark a New Round of Competition!
“Difficult to Obtain Masks”? Which Electronic Companies Are Starting to Cross-Produce Masks?

Xiaomi’s New Year First Battle! All Mi 10 Models Support Wi-Fi 6! Wi-Fi 6 Enters the Popularization Year

Overall Solution for R&D Testing of IoT Products

Keysight U5855A Infrared Thermal Imager Helps You Develop Sharp Eyes

Frequent Acquisitions of WiFi Chips, 2020 Will Spark a New Round of Competition!

“Difficult to Obtain Masks”? Which Electronic Companies Are Starting to Cross-Produce Masks?

Related posts

Leave a Comment Cancel reply