The era of Moore's Law is over, where is the semiconductor industry going?

It is reported that the international semiconductor technology roadmap to be published next month will no longer target Moore's Law. The global semiconductor industry will formally endorse a problem that has been discussed for a long time: Moore's Law, which has been driving the IT industry since the 1960s, is coming to an end. Where will the semiconductor industry officially abandon Moore's Law?

Nature published a recent article to discuss this.

The following is the main content of the article compiled:

Moore's Law can be said to be the most important law of the entire computer industry. It is actually a prediction: the number of transistors in a microprocessor will double every two years - meaning that the processing power of the chip is doubled. This exponential growth led to the transformation of large home computers in the 1970s into more advanced machines in the 1980s and 1990s, and then gave birth to high-speed Internet, smartphones and now car networking, smart refrigerators and auto tuning. Thermostat and so on.

This seemingly natural process is largely the result of human intentional control. Chip manufacturers are interested in following the trajectory of Moore's Law prediction: software developers' new software products are increasingly challenging the chip processing capabilities of existing devices, and consumers need Updated to a higher-profile device, device manufacturers rush to produce next-generation chips that meet processing requirements.

Since the 1990s, the semiconductor industry has released a blueprint for industry R&D planning every two years, coordinating hundreds of chip manufacturers and suppliers to follow Moore's Law. Such a strategy is sometimes called "more". More Moore, due to the existence of this blueprint, the entire computer industry followed Moore's Law step by step.

But now, this development trajectory has come to an end. As more and more silicon circuits are integrated in the same small space, the heat generated is also increasing, and the speed of doubling the processing capacity in the past two years has slowly declined. In addition, there are more and more problems that are slowly emerging. Today's top chip manufacturers have circuit accuracy of 14 nanometers, which is smaller than most viruses. However, Paolo Gargini, president of the Global Semiconductor Industry R&D Planning Blueprint Association, said: "By 2020, at the fastest rate of development, our chip line can reach 2-3 nanometer level, but in this Only 10 atoms can be accommodated in the class. Can such a device be called a 'device'?"

I am afraid not. At that level, the behavior of electrons will be limited by quantum uncertainty, and transistors will become unreliable. In this prospect, although there have been countless studies in this area, there is still no way to find new materials or technologies that can replace today's silicon technology.

The industry research blueprint released next month will be unprecedentedly centered on Moore's Law. On the contrary, the new strategy may be “More than Moore”: different from the previous development of chips and software. The future development of the semiconductor industry will first look at software—from mobile phones to supercomputers to cloud-based data centers—and then look at chips that support the processing power of software and applications that need to be supported, thanks to new computing devices. It is becoming more and more mobile, and there may be a new generation of sensors, power management circuits and other silicon devices in the new chip.

This shift in the situation has also changed the semiconductor industry to no longer unite around Moore's Law. “Everyone is not sure what the new research plan blueprint means,” said University of Iowa computer scientist Daniel Reed. The Semico nductor Industry Association (SIA), based in Washington DC, represents all US semiconductor companies and has indicated that it will no longer participate in the charter of the global semiconductor industry research blueprint, but will decide on its own development progress.

Although Moore's Law has gone to dusk, it does not mean that the semiconductor industry has stopped developing. Daniel Reid compares it to the aircraft manufacturing industry: "The current Boeing 787 is no faster than the Boeing 707 of the 1950s - but the two models are much worse, and the Boeing 787's innovation is reflected in Other places, such as full electronic control, carbon fiber fuselage, etc., are also true in the computer industry. Innovation will continue, but it will be reflected in smaller and more complex places."

The birth of Moore's Law

Prior to the publication of the famous paper in 1965, Gordon Moore was the R&D director of Fairchild Semiconductor, Inc. in San Jose, Calif., who had predicted home computers, electronic watches, self-driving cars, and "personal "Mobile communication device" - the birth of the mobile phone, but the 1965 paper on the predictions that were later called "Moore's Law" really made him famous. The core of this paper is about the development of the future computer industry. Based on the understanding of Fairchild and other semiconductor companies, Moore expects to double the number of transistors and other electronic components per chip per year.