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

This article comes from the official account of Wechat: SF Chinese (ID:kexuejiaodian), written by Peter Bentley, translated by Li Shijie

Today's computer processors are much more powerful than they were 50 years ago. The computing speed of the processor doubles roughly every two years.

This doubling effect, known as Moore's Law, is named after Intel co-founder Gordon Moore, who predicted the trend as early as 1965.

If the car's top speed follows the same growth trend since 1965, British driver Lewis Hamilton (Lewis Hamilton), the seventh World Formula one Championship (F1) champion, will be speeding along the track at more than 11 billion miles per hour.

Approaching the tipping point for the computer industry, Moore's prediction turned into some kind of self-fulfilling prophecy.

The chipmaker was inspired by Moore to achieve the performance predicted by Moore's law. They did the same: shrinking electronic components, adapting smaller and smaller chips, and speeding up the rate of interaction between electronic components.

This highly integrated circuit makes the performance of the processor more and more powerful, and the computer changes the world to a great extent. From food distribution to transportation, digitization involves almost every aspect of life, and scientists and engineers have also developed new technologies that previous processors could never achieve, such as social media, online games, robots, augmented reality (AR) and so on.

The continuous progress of chip technology makes these changes possible. But people have become accustomed to rapid technological innovation, and software developers believe that Moore's law still applies.

However, today's Internet world produces a huge amount of data every day, which gives rise to a huge demand for data storage, and the larger the amount of data, the greater the data analysis capability.

But the story of the rapid development of chips will not last forever. We are approaching a tipping point, because the size of the transistor has reached the extreme, and if we continue to shrink the transistor, we will encounter the limitations of physical laws.

For example, the A11 chip released by Apple in September 2017 was one of the best chips at the time, with an 87.66mm square chip containing 4.3 billion transistors.

If it is smaller than the Apple A11 chip, which means that the transistor will become smaller, then the transistor will be disturbed by the quantum effect (electron transition).

In addition, because the space is too small, the internal structure of the chip is arranged, and it is very difficult to control the electrical performance of the chip. If there are too many transistors, the working speed will be accelerated, and if the internal impedance of the chip is too small, the temperature of the chip will be too high, and if it cannot be cooled down immediately, the chip will ignite spontaneously.

Apple released the A16 chip in 2022, which is significantly better than the A11, but also closer to the physical limit.

A16 contains about 16 billion transistors, because the size is too small, the quantum effect is already very obvious. In order to prevent the emergence of electron tunneling effect, the chip manufacturing company improved the process and covered the transistor with multiple layers of silicon.

By constantly improving the process, the number of transistors on the chip can continue to increase in the next few years, but even so, the transistor cannot be infinitely smaller, because the diameter of the silicon atom is 0.2 nanometers, no matter how small the transistor is. It can't be smaller than a silicon atom.

Therefore, in order to further improve the performance of the computer, we can only find a way to bypass the quantum effect.

Alternative strategy: dedicated chipmakers have been doing their best to solve these problems for decades. The clock rate of the processor (the basic running speed of the computer) is increasing every year.

In 1991, Intel's i486 processor clock rate was 25MHz:1998 year, Pentium Pro processor clock rate was 200MHz Pro 2008, and Pentium 4 processor clock rate was 3.8GHz.

But this is the fastest speed that can be achieved. Since then, manufacturers have had to use multi-core processors so that they can work in parallel and run faster-first dual-core, then quad-core, eight-core, 16-core, and so on.

Continuing to improve chip performance at the rate predicted by Moore's Law has been so difficult and costly that almost all chipmakers have abandoned the race. It is no longer cost-effective to continue to move in this direction, so there are significantly fewer laboratories engaged in the research and development of cutting-edge chip manufacturing processes. The era of Moore's Law is coming to an end.

Instead, manufacturers focus on developing dedicated chips to speed up specific types of algorithms. For example, the most common graphics processor.

The GPU was originally developed to execute multiple algorithms in parallel to help computers quickly process the image data of games. GPUs have now evolved into general-purpose processors for data analysis and machine learning. Chipmakers have also introduced application-specific integrated circuits (ASIC), such as Google's tensor processors.

Tensor processor is a parallel working circuit composed of 256 chips, which is specially developed to improve the operation speed of deep neural network.

However, the end of Moore's Law should not be seen as the end of progress.

The era of creating new possibilities has come, and scientists are developing new computer architectures and technologies in the real sense, and are devoting more energy to materials science, looking for materials that can replace silicon to make transistors.

These new technologies show that although the computing speed of computers may no longer grow exponentially, they will still improve rapidly, and data processing will be carried out in entirely new ways.

The original article is published in Science focus 202212.

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.