Shulou(Shulou.com)11/24 Report--

The literal meaning of computing power, which we all understand, is computing power (Computing Power).

The so-called "calculation", we can have a variety of definitions.

In a narrow sense, it is the process of performing operations on mathematical problems, such as the completion of "1-1-1-1". Or the process of reasoning from the Goldbach conjecture.

In a broad sense, the definition is more macroscopical, and the process of processing information and getting results can be called "calculation".

Obviously, the difference between the narrow definition and the broad definition is mainly the difference in the content of calculation. The ability to complete the computing process can be called "computing power".

In fact, human thinking is one of the most common computing processes.

Apart from sleeping and in a daze, we think all the time. We observe, perceive and collect external information through our facial features. Then, with the help of the brain, the information is processed (that is, thinking). Finally, draw conclusions, make judgments, and take action.

In this process, the brain is our computing tool. And the brain's ability to think is to calculate power. The faster the brain thinks, the more powerful it is.

Computing is a way for human beings to solve problems.

In the long history, human beings have encountered many problems, which need to be solved by calculation. These computing tasks cannot be accomplished by the brain alone as a "native" computing tool. As a result, human beings have invented a lot of computing tools and methods to meet the computing needs. Such as abacus, abacus, slide rule and so on.

In the 1940s, with the continuous accumulation of technology, the electronic computer was born and the information technology revolution officially began.

The early computer was actually a large calculator, mainly used for complex computing tasks in the military field (such as ballistic calculation). Its performance is not strong, and its size and power consumption are huge. Later, transistors were invented to replace vacuum tubes, and the problems of size and power consumption were gradually solved.

In 1958, integrated circuits came out, officially creating the chip era. The chip has a large number of electronic components (such as transistors, resistors, capacitors, etc.), which can execute operational instructions. In recent decades, under the domination of Moore's Law, the number of transistors on the chip is increasing, and the performance is also improving.

With the blessing of chip capability, computers become more and more powerful and smaller and smaller, which eventually gives birth to PC and a prosperous IT software and hardware ecology. Computers began to enter the family and industry, and eventually become the most important computing tool for human beings.

We use computers in various fields to run programs, solve problems and improve efficiency. The more advanced the chip manufacturing process, the more transistors, the more powerful the computing power, and the faster and better the problem can be solved.

SoC chip for 5G mobile phone nowadays, chip has become synonymous with computing power. When we talk about computing power, we are actually talking about the computing power of the chip.

Generally speaking, the industry tends to refer to chip technologies and capabilities such as CPU and GPU as computing power in a narrow sense. Storage technology related to memory and hard disk is called storage power. The software technology, including operating system, database, middleware, application and so on, is called algorithm.

The computing power in a broad sense includes not only the arithmetic power in the narrow sense, but also the survival power and algorithm.

Cutting-edge concepts such as cloud computing, big data, artificial intelligence and blockchain all belong to the application of computing power. In other words, everything related to information technology can be generally referred to as the field of computing.

We also need to note that chips are the core of computing power, while mobile phones, watches, PC and other terminals with chips, as well as servers and other devices, are the carriers of computing power. Data centers with a large number of servers, as well as computing clusters, can also be called computing platforms. They are the main forms of arithmetic.

The value of █ computing power the role of computing power is to complete the computing task.

As we all know, the operation of the computer hardware system and the execution of the program software are supported by countless computing tasks. Therefore, the computing power provided by the chip is the power source for the normal operation of the whole system.

After years of popularization, information technology has spread to every corner of our work and life. Various IT systems support the development of the whole society. Computing power supports all IT systems, while IT systems support the whole society. From this point of view, it is not too much to regard arithmetic as the cornerstone of social development.

In terms of life, our clothing, food, housing, transportation, entertainment and leisure are inseparable from mobile phones and mobile Internet. Our mobile phone is the chip inside to provide computing power, so that it has rich functions and smooth speed.

The digital e-commerce we visit, the online games we play, and the movies and videos we watch are all based on the services of Internet service providers. These services are built on data centers and servers, and the chips provide computing power. The stronger the power of calculation, the better the service experience, and our life will be more convenient and happier.

In terms of work, various industries are now promoting digital transformation, combining advanced IT technology and communication technology with traditional industries.

Digitalization is a further extension of informatization. In the past, informationization only introduced IT technology into some specific services. And digitalization is the transformation for the whole enterprise. Including organizational structure, business processes, business models and work scenarios, are the objects of digital transformation.

The purpose of digitization is to improve production efficiency, reduce costs and enhance the comprehensive competitiveness of enterprises.

Whether it is informationization or digitalization, computing power is behind it. The stronger the computing power, the stronger the ability of the system, the greater the improvement and the more benefits.

Some enterprises have developed in the direction of intelligence on the basis of informationization and digitalization. In this way, the improvement of efficiency will be greater, forming a "generation gap" level of technical advantage. In the increasingly fierce market competition in the future, this advantage can determine the life and death of enterprises.

Now there is a popular saying in the industry that all business models are closer to the direction of "mining data value".

Data is regarded as the most valuable resource and is a rich mine. And calculus is seen as a tool for digging the mine. Through computing power to process the data, we can tap the huge data value and create wealth.

The process of mining data value is subdivided into four links: generating data, transmitting data, storing data and calculating data. Computing power (information technology) and linkage (communication technology), working together, can accomplish this process:

First of all, we use sensors, cameras and other devices to collect information from the physical world and convert it into digital bits. Then, it is transmitted and transported through 5G, Wi-Fi, optical fiber and other communication technologies. These digital bits are stored in storage media such as hard drives and then handed over to the chip for calculation. The calculated results are applied to decision-making and control.

With the blessing of artificial intelligence technology, the protagonist of decision-making and control may no longer be us, but AI agents.

See, the role of computing power is undoubtedly revealed in the process of data value mining. Without strong math, you can't finish this promising job.

The important value of computing power is also reflected in the level of national competitiveness.

The computing power determines the development speed of the digital economy and the height of the development of social intelligence. According to the data jointly released by IDC, Tidal Wave Information and the Global Industrial Research Institute of Tsinghua University, for every 1 point increase in computing power index, the digital economy and GDP will increase by 3.5 ‰ and 1.8 ‰ respectively.

There has been a significant positive correlation between the scale of computing power and the level of economic development in countries all over the world. The greater the scale of a country's computing power, the higher the level of economic development.

It is no exaggeration to say that arithmetic has become an important part of national competitiveness.

The classification of █ computing power serves the whole society. On the other hand, there are differences in social demand for numeracy. These arithmetic demands come not only from the field of consumption (mobile Internet, drama chase, online shopping, taxi hailing, O2O, etc.), but also from the field of industry (industrial manufacturing, transportation and logistics, financial securities, education and medical care, etc.). There are also from the field of urban governance (smart city, one card, city brain, etc.).

Different computing applications and requirements have different algorithms. Different algorithms have different requirements for the characteristics of computing power.

Today, we divide computing power into three categories, namely, general computing power, intelligent computing power and supercomputing power.

The general computing power is based on the output computing power of CPU (Central Processing Unit, CPU). There is an instruction set inside CPU to guide and optimize the operation to ensure the reliable operation of CPU.

According to the instruction set architecture, CPU can be divided into x86 architecture and non-x86 architecture. Everyone is familiar with the X86 architecture, which was first developed and led by Intel for a long time. It has a good ecology and a high market share. There are many types of non-x86 architectures, which are rising rapidly in recent years, mainly x86, ARM, MIPS, Power, RISC-V, Alpha and so on.

The intelligent computing power is mainly based on the output computing power of GPU (Graphics Processing Unit, graphics processor), FPGA (Field ProgrammableGate Array), AI (Artificial lntelligence) chips and so on. GPU, in particular, can be said to be hot at present, and it is difficult to get a card.

The supercomputing power is mainly based on the computing power output of the supercomputer. It uses the centralized computing resources of multiple computer systems working in parallel, and uses a special operating system to deal with extremely complex or data-intensive problems. It is mainly used in cutting-edge scientific research, national defense and other sophisticated fields. The price is extremely expensive, but the performance is also extremely strong.

In the data center, computing tasks are also divided into basic general computing and HPC high performance computing (High-performance computing).

HPC calculations continue to be subdivided into three categories, namely:

Scientific computing: physical chemistry, meteorological environmental protection, life science, petroleum exploration, astronomical exploration and so on.

Engineering calculation: computer-aided engineering, computer-aided manufacturing, electronic design automation, electromagnetic simulation and so on.

Intelligent computing: artificial intelligence computing, including: machine learning, deep learning, data analysis and so on.

Scientific computing and engineering computing you should have heard of, these professional scientific research areas of data production is very large, the demand for computing power is extremely high.

Take oil and gas exploration as an example. Oil and gas exploration, to put it simply, is to CT the surface. A project down, the original data is often more than 100TB, or even more than 1 PB. Such a huge amount of data needs a huge amount of computing power to support.

Intelligent computing is very popular in recent years, and it is the development direction that the whole society pays close attention to. Driven by the AIGC model, various industries are making great efforts to develop intelligent computing, which has a great demand for intelligence.

According to the type of computing power, the data center we usually talk about is usually divided into general data center, intelligent center and supercomputing center.

The Internet service that we usually use mainly comes from the general data center. The intelligent computing center is a data center that specializes in intelligent computing. The supercomputing center specializes in a variety of large-scale scientific and engineering computing tasks, all of which are supercomputers such as Tianhe-1.

In the arithmetic unit, there is now a more detailed division according to the division of tasks. In addition to the CPU and GPU just mentioned, TPU, NPU and DPU have emerged one after another in recent years, which are also special computing units with specific computing tasks.

The trend of █ Computing Power Computing Power and Relay Force are important components of digital productivity. In recent years, with the continuous deepening of information, digitization and intelligence, the whole society has a strong demand for computing power.

Driven by demand, the development of computing power has also shown the following trends:

With the continuous growth of computing demand, the arrival of the era of the Internet of everything, the introduction of a large number of intelligent Internet of things terminals, the promotion of the digital transformation of the industry, coupled with the landing of AI intelligent scenes, will produce unimaginable amounts of data. These data will further stimulate the demand for computing power.

According to Roland Berger's forecast, from 2018 to 2030, the demand for computing power in autopilot will increase 390 times, the demand for smart factories will increase 110 times, and the demand for computing power per capita in major countries will increase 20 times from less than 500 GFLOPS today to 10000 GFLOPS in 2035.

According to the forecast of Chaochao artificial Intelligence Research Institute, the global computing power will reach 6.8 ZFLOPS by 2025, a 30-fold increase compared with 2020.

In order to meet such a huge demand for numeracy, efforts need to be made in the following aspects.

First of all, constantly improve the process of the chip itself, integrate more transistors, and improve the single point computing power of the chip. After decades of development, Moore's Law has gradually moved towards the physical bottleneck, the chip process is close to 1nm, the space that can be improved is very limited, and the price will be greater.

Second, build a large number of computing infrastructure, such as data centers. Through large-scale, to meet the needs of the whole society.

Finally, through the new computing service models such as eastern and western computing and computing network, we can strengthen the effective utilization of computing power, so as to appropriately alleviate the pressure of the growth of computing demand.

When the previous article introduced the classification of arithmetic power, it was mentioned that arithmetic power can be divided into three types: general arithmetic power, intelligent arithmetic power and super arithmetic power.

In fact, this classification has only begun to take shape in recent years. General-purpose computing power plays a leading role in the demand for computing power. However, with the rapid development of artificial intelligence technology such as AIGC large model, the proportion of intelligent computing power begins to rise rapidly.

According to the "China Comprehensive Computing Power Index (2023)" released by the China Institute of Information and Communication, at present, the scale of general computing power accounts for 74%; the scale of intelligent computing power accounts for 25%. Although the proportion of smart computing power is still less than that of general computing power, it is growing very fast, increasing by 45% over the same period last year. This growth rate is also higher than the overall growth rate of computing power.

In other words, the development of the AIGC model has significantly promoted the demand for intellectual computing power. The overall structure of the field of computing is changing, and the demand for intelligent computing is showing explosive growth.

This also means that the proportion of the construction of intelligent computing centers will increase significantly in the follow-up construction of computing infrastructure. The development of smart computing industry will also enter a golden period of development.

Computing services were ubiquitous in the early days of mobile mainframes, and computing provided services in a centralized manner. After the PC appears, the computing power begins to enter the user side. The popularity of mobile phones and the Internet in the 1990s broke the fixed space of computing and began to "move".

With the continuous iterative upgrading of mobile chips, the computing power of users' mobile phone terminals is growing, which is almost comparable to that of PC chips.

On the other hand, based on the development of mobile communication technologies such as 5G and Wi-Fi, the Internet of everything begins. There are more and more types of terminals, and they all have large or small computing power and the ability to calculate at the end.

After the rise of cloud computing, computing began to cloud and distribute. Edge computing appears, and computing power sinks from the cloud to all levels of the communication network.

All this marks the beginning of the flow of computing power, all over the corner of the cloud tube. This is the generalization of computing power.

The computing power network mentioned just now is actually an embodiment of the ubiquity of computing power.

Green and low-carbon computing facilities support the development of the whole society, but the energy consumption problems it brings are becoming more and more obvious.

According to the data, the total electricity consumption of the national data center in 2021 is 216.6 billion kilowatt-hours, accounting for 2.6 percent of the country's total electricity consumption, equivalent to the annual power generation of the two three Gorges hydropower stations and the total electricity consumption of 1.8 Beijing areas.

Such terrible power consumption has not only caused great pressure on us to achieve the goal of "double carbon", but also seriously affected the sustainable development of the world economy. Therefore, finding ways to reduce the energy consumption of computing power has become the key research direction of the whole industry.

There are many ways to realize the green and low-carbon computing. Through basic theoretical research, material process upgrading, R & D and technological innovation, the power consumption control and improvement of computing infrastructure is the most effective means of energy saving and emission reduction from the source.

In addition, increasing the proportion of renewable energy and reducing the use of fossil energy is also the key to the green development of computing power.

At present, the research on energy saving and emission reduction in all aspects of computing power has achieved preliminary results. With the green development of computing power, the overall prospect is very optimistic.

According to the Green Development 2030 report, the energy efficiency of the global digital infrastructure will increase by 100 times by 2030, with renewable energy accounting for more than 50 per cent of electricity generation and industry digital penetration reaching 50 per cent.

The exploration of new computing power accelerates the continuous growth of the demand for computing power, which puts great pressure on the traditional semiconductor chip technology. After the semiconductor process enters the bottleneck, more and more experts begin to study the new computing technology theory, such as quantum computing, optical computing, brain-like computing and so on.

Quantum computing can surpass the computing power of classical computers by using quantum superposition states and quantum entangled states. Photon computing (also known as optical computing) is a computing method that uses light waves for data processing, data storage or data communication. Brain-like computing realizes intelligent learning and decision-making ability by simulating the neural network and synaptic connections of the brain.

At present, these new fields of numeracy are in the research stage, and some achievements have been made, but they are also faced with many difficulties.

Once there is a real breakthrough in these fields, the traditional arithmetic framework will be completely subverted, and human society will enter a new stage of development.

All right, that's all for today's article. Thank you for your patience in reading!

This article comes from the official account of Wechat: fresh Jujube classroom (ID:xzclasscom), author: Xiaozaojun

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.