Has Moore's law come to an end?

Focus on + Star sign public Number , Don't miss the highlights

Direct sources | Semiconductor industry observation

source ：https://mp.weixinbridge.com/mp/wapredirect?url=https%3A%2F%2Finterestingengineering.com%2Ftransistors-moores-law

For at least the past decade , People have been talking about the end of Moore's law and what impact it will have on modern society . 

since 1947 Since the invention of , Transistors continue to power the modern world , The number of transistors packaged on silicon chips has steadily increased in density , This makes the past 70 Computing power has grown exponentially over the years .

However , A transistor is a physical object , And is a purely physical object , Like all other physical objects, it is bound by the laws of physics , This means that there are physical limitations to the size of the transistor . 

But in Gordon · Mole （Gordon Moore） When making his famous prediction about the growth rate of computing power , Nobody really thought about nanoscale transistors . 

But as we enter 21 The third decade of the 20th century , Our dependence on packaging more transistors in the same amount of silicon is breaking through the boundaries of physical possibilities , Many people worry that the pace of innovation that has made us accustomed may come to an abrupt end in the near future .

The history of transistors

A transistor is a semiconductor , There are usually at least three terminals that can be connected to the circuit . Usually , One of the terminals is responsible for controlling the current through the other two terminals , This allows fast switching in digital circuits .

Before the advent of transistors , This fast circuit switching is accomplished by using a thermionic valve , Thermionic valves are often called old vacuum tubes . 

These vacuum tube triodes are much larger than transistors , And it needs more power to run . Unlike transistors , They are not “ solid state ” Components , This means that they may fail during normal operation , Because they rely on the movement of electrons flowing in the tube to conduct electron currents .

This means that electronic devices based on vacuum tubes are large 、 High temperature and high operation cost , Because they need regular maintenance to replace electronic tubes that have failed for some reason , So that the whole electronic machine stops running .

The transistor is in AT&T The Bell Laboratories of John Bardeen and Walter Houser Brattain stay William Shockley Under the supervision of “ The invention ” Of . Even before that , The concept of transistor has existed for about 20 year —— But until Bell Labs finished the work , To establish the working model of the transistor .Shockley stay 1947 The design was improved in 1948 Bipolar junction transistor of , It is this implementation that is implemented in 1950 It was first put into mass production in the s . 

The next big leap is silicon surface passivation , It allows silicon to replace germanium as the semiconductor material for transistors , Later it was used in integrated circuits . 

1959 year 11 month , Bell Labs Mohamed Atalla and Dawon Kahng Invention of metal oxide semiconductor field effect transistor (MOSFET) , It is better than Shockley Bipolar junction transistors consume much less energy , And more scalable .

MOSFET It is still the main transistor used today , And as a separate unit , It is the most manufactured device in human history . because MOSFET It can be made smaller and smaller , More and more transistors can be made into integrated circuits , So as to realize more and more complex logical operations .

To 1973 year ,RCA Executive vice president of research and engineering William C. Hittinger Brag about “ In silicon only a few millimeters wide ‘ chip ’ Put it on 10,000 Multiple electronic components .” Today's transistor density far exceeds these early advances by orders of magnitude .

Gordonmoore accidentally invented Moore's law

Gordon Moore Not a household name , But his works are found in almost every home and office in the industrialized world . Although he will continue to be president of Intel Corporation , And eventually become its Honorary Chairman , But when he was 1965 When describing what we now call Moore's law in , He is not so respected .

As an electrical engineer , Moore used to work in the Shockley Semiconductor Laboratory Department of Beckman instruments , Then Shockley himself led . But later , Shockley Several employees of , Even some of his disciples , All the Shockley Of the leadership , So they went to 1957 It was established independently in Fairchild Semiconductor, This is one of the most influential companies in history .

As the R & D director of Fairchild Semiconductor , Moore is a natural person who asks about the current situation of the industry , therefore 1965 year 《 Electronics 》 The magazine asked Moore to predict the development direction of the semiconductor industry in the next decade . have a look Fairchild The speed of innovation , Moore just extrapolated forward in time .

In the years since Fairchild began manufacturing semiconductors , The cost of producing components is reduced , The size of the component itself is reduced by about half every year . This makes Fairchild The same number of integrated circuits are produced every year , But the number of transistors is twice that of the previous year .

2013 year , Gordon · Mole

“ I didn't expect this estimate to be accurate ,” Moore in 1995 In, he wrote .“ I just want to convey an idea , That is, this is a technology with a future , In the long run , It is expected to make a considerable contribution .”

“ I think this is really a great achievement for the industry . Keep this exponential growth 35 year , And the density has increased by thousands , It is really difficult to predict with confidence ,” Moore added .

Moore's prediction remained basically stable for about a decade , Moore then revised his estimate to double the transistor density every two years .“ I have never been able to see future generations [ semiconductor ] Any details beyond . However , It's amazing , Generation after generation has been springing up , Keep us on the same slope ,” Moore wrote .“ The current forecast is , It won't stop soon .” This is in 1995 Years may be right , But soon Moore's law will begin to break through the boundaries of Physics , It will begin to face the challenge of survival .

Why is Moore's law in trouble ？

Standing on the 2022 See Moore's law in , The problem is that the size of the transistor is now very small , So that we can't do more to make them smaller . According to Taiwan semiconductor manufacturing company 2024 Production roadmap for , Transistor gate （ The part of a transistor through which electrons flow as current ） The width of is now close to 2 nanometer .

The width of a silicon atom is 0.2 nanometer , This makes 2 The gate length of nanometers is about 10 A silicon atom . On these scales , As various quantum effects play a role in the transistor itself , It becomes more and more difficult to control the flow of electrons . For larger transistors , Crystal deformation on the atomic scale does not affect the overall flow of current , But when you only have about 10 Atomic distances can be used , Any change in the underlying atomic structure will affect the current flow through the transistor . Final , Transistors are approaching the point where we can make them as small as possible and still work . The way we have been building and improving silicon chips is about to go through the last iteration .

Moore's law has another potential pitfall , That is simple economics . Reducing the cost of transistors is not like 1960 The age is so low . at best , It has declined slightly from generation to generation , But diseconomies of scale are beginning to affect manufacturing . When the demand for semiconductor chips started , The engineering capacity to produce chips is expensive , But at least it's available . As the demand for smart phones, satellites and the Internet of things soars , There is not enough capacity to meet this demand , This leads to higher prices at every step of the supply chain .

what's more , When the number of transistors is doubled , They also generate more heat . For many companies that are the largest buyers of the most advanced processing chips , The cost of cooling large server rooms is becoming more and more unbearable . As businesses try to extend The life and performance of its existing equipment to save money , The chipmakers responsible for implementing Moore's law have reduced their R & D revenue —— And R & D itself has become more expensive .

If there is no additional benefit , It is even more difficult to overcome all the physical obstacles to further shrinking the transistor . therefore , Even if physical challenges don't end Moore's law , But the lack of demand for smaller transistors will almost certainly end .

So what are we doing ？

Um. , This is the current trillion dollar problem . In the past 70 Mid year , We have experienced unprecedented technological progress , So at this point , Almost every industrialized society regards rapid technological progress as a given condition . 

Why did you suddenly stop ？ What would it feel like ？30 Years have the same iPhone What does it mean ？ obviously , We can simply treat it as a society . our DNA Nothing in this book requires us to have a new one every two to three years iPhone, Have a brand new computer every five years . We are just used to the speed of progress , If this speed changes , We will also adapt ourselves to this situation . 

After all , Humans have had computers for less than a century , About the time we existed on this planet as a species 1/250,000. We will find a way to get through such arduous hardships .

perhaps , We can look forward to the end of Moore's law with excitement and anticipation . After all , Adversity is the mother of invention . In the past 70 In the year , We've been trying to figure out how to shrink transistors more and more , Now the road of innovation has come to an end . 

This is definitely not the only way forward , If we don't focus all our efforts on shrinking the transistor , We can focus on other areas , Discover new breakthroughs , by comparison , The invention of the transistor may seem mediocre . Before exploring these new innovative approaches , We won't know , The end of Moore's law may be the signal we need , It's time to start looking for new engines of progress .

Moore's law is dead ！ Long live Moore's law ！

in the final analysis , Moore's law was not a “ The laws of ”, It's more like a self fulfilling wish . We expect transistor density to double every year , And then double it every two years , So we are looking for ways to accomplish this task .

Whatever comes next , Whether it's quantum computing 、 Machine learning and artificial intelligence , Even something we don't even have a name for , We will all find a new desire to push this innovation forward .

in the final analysis , Our fascination with Moore's law has never been the density of transistors . Most people who have heard of Moore's law cannot even begin to explain the meaning of transistor density , Not to mention how interlocking transistors form logic circuits or how smart phones in your pocket work （ Or even 1970 The pocket calculator of the s ） . For most of us , Moore's law is always about our expectations of progress , And this is largely up to us .

Moore's law may have come to an end , But if we really want it , We'll find a new Moore's law .

------------ END ------------

Pay attention to the reply of the official account “ Add group ” Join the technical exchange group according to the rules , reply “1024” See more .

Click on “ Read the original ” See more sharing .