Taking the first step, Huawei has developed its own Kirin NPU chip. How much do you know about the chip?

time:2023-03-24 09:16:01 source:scripttoolbox.com author:Small parts
Taking the first step, Huawei has developed its own Kirin NPU chip. How much do you know about the chip?

In terms of chips, Huawei has always insisted on independent research and development. Among them, Huawei's mobile phone chips already have Kirin series chips, and communication chips have Balong series chips. In the field of server chips, Huawei also has Kunpeng series chips. In addition, Huawei has also developed RF chips, router chips, etc. So what is an NPU chip? The embedded neural network processor (NPU) adopts the "data-driven parallel computing" architecture, and is especially good at processing massive multimedia data such as video and images. Chips equipped with neural network processors are used in surveillance cameras, and the cameras are upgraded from "eyes" to "eyes with brains". The popular point is to add a chip to the camera that can intelligently identify items, turning the camera into an "eye". Huawei's self-developed NPU chip is HiSilicon's NPU. This HiSilicon NPU will be placed on the Mate50 mobile phone to enhance the overall performance of the mobile phone and reasonably control the power consumption of the mobile phone. The specific performance will have to wait for the Mate50 release. Looking forward to it? NPU, GPU, CPU, etc. are all kinds of processing units. How many kinds of chips do you know? If I have only heard of what I mentioned above, let me give you some popular science.

1.APU (Audio Processing Unit)

The Chinese name is Acceleration Processor, which is the fusion future launched by AMD in 2011 Idea product. For the first time, it integrates the processor and the independent graphics core on one chip, co-computing and accelerating each other, so that tasks can be flexibly distributed between the CPU and GPU, improving efficiency. This is the statement on AMD chips. Another common APU sound processor is the processor on the sound card, which can be used to process sound data. As the processing of surround sound in modern computers (especially home systems, game enthusiasts) has become more and more The more important, the need for a dedicated sound core processor.

2.BPU (Brain Processing Unit)

BPU (Branch Processing Unit) is an area in the CPU used for branch processing. In addition, the BPU is also an AI chip developed by an autonomous driving chip and solution company. Based on the innovative artificial intelligence-specific computing architecture BPU, Horizon has successfully tape-out and mass-produced China's first edge artificial intelligence chip--Journey, which focuses on intelligent driving. The three-generation BPU (Brain Processing Unit, an efficient artificial intelligence processor architecture independently designed and developed by Horizon) architecture.

3.CPU (central processing unit)

Everyone is familiar with CPU, that is, the central processing unit (CPU, central processing unit) as the calculation and control of the computer system The core is the final execution unit of information processing and program operation, and the core of computing and processing data.

4.DPU (Deep-Learning Processing Unit)

The concept of DPU was first proposed by Shenjian Technology, established in 2016, and its products are based on FPGA The processing unit has industry-leading machine learning capabilities, focusing on neural network pruning, deep compression technology and system-level optimization. Acquired by Xilinx in 2018. At the same time, DPU is the abbreviation of Data Processing Unit. It is a major category of newly developed special-purpose processors. It is the third important computing power chip in data center scenarios after CPU and GPU. It is a high-bandwidth, low-latency chip. , Provide computing engine for data-intensive computing scenarios.

5.EPU (Emotion Processing Unit)

The word EPU is basically known to few people. EPU is proposed by Emoshape company. Emoshape is a company dedicated to providing technology that can teach intelligent machines to interact with humans. Based on microcontroller (MCU) designs, EPUs enable nearly limitless cognitive processes in artificial intelligence (AI), robotics, and consumer electronic devices to generate true emotional responses. EPU is an extension of Ekman's theory, which uses the theory of emotional evolution to identify 12 major emotions. The main function of the EPU is to allow artificial intelligence and robots to have emotions and talk to people, which is one of the AI ​​chips.

6.FPU (Floating Processing Unit)

Module for floating-point operations

7.GPU ( Graphics Processing Unit)

Graphics processor (English: graphics processing unit, abbreviation: GPU), also known as display core, visual processor, display chip, is a kind of specialized in personal computers, workstations, game consoles and some mobile A microprocessor that performs image and graphics-related operations on devices (such as tablet computers, smartphones, etc.). GPUs are widely used in embedded systems, mobile phones, personal computers, workstations and video game solutions. Modern GPUs are very efficient for image and graphics processing because GPUs are designed with a highly parallel architecture which gives them an advantage over general-purpose CPUs in parallel processing algorithms on large blocks of data. Therefore, GPU has great advantages in tasks such as image processing. Combined with application scenarios such as autonomous driving, GPU has become the new favorite of the chip industry.

8.HPU (Holographic Processing Unit)

Holographic Processing Unit (HPU) can work with CPU and GPU A working coprocessor chip, announced by Microsoft on January 21, 2015. This is a custom chip that can process and interact with data streams from different sensors and Intel Atoms.

9.IPU (Intelligence Processing Unit)

IPU (Intelligence Processing Unit) is a revolutionary technology for AI computing Architecture, today, IPU has achieved results in the fields of finance, medical care, telecommunications, robotics, cloud and the Internet. It can be used for extreme scale-out of machine intelligence, offering unprecedented AI computing possibilities, complete flexibility and ease of deployment.


With the rapid development of technology today, the concept of robots and autonomous driving has also emerged in the past two years. They are all inseparable from the blessing of the A chip. Therefore, a lot of AI chips have been developed. The above are just some of the PUs I introduced. There are many other AI chips such as: TPU (Tensor Processor), PPU (Physical Computing Processor) and so on. Today, my country's chip development business is also booming. The mid-to-low-end chips (14nm-90nm) have made great achievements, but the development of high-end chips still needs to be improved, and the goal of self-sufficiency rate reaching 70% in 2025 is still a long way off. Now many domestic manufacturers have launched their own chips, such as Xiaomi's Surging series, Huawei's HiSilicon and Kirin series, domestic self-developed chips Ziguang Zhanrui, Cambrian Siyuan series chips and so on. Hope our Chinese core will become stronger and stronger!

(Responsible editor:Monitor)

Related content