How Moore's Law Works

by Jonathan Strickland

Gordon Moore at Intel's headquarters

AP Photo/Paul Sakuma
Introduction to How Moore's Law Works

There's a joke about personal computers that has been around almost as long as the devices
have been on the market: You buy a new computer, take it home and just as you finish
unpacking it you see an advertisement for a new computer that makes yours obsolete. If you're
the kind of person who demands to have the fastest, most powerful machines, it seems like
you're destined for frustration and a lot of trips to the computer store.

While the joke is obviously an exaggeration, it's not that far off the mark. Even one of today's
modest personal computers has more processing power and storage space than the famous
Cray-1 supercomputer. In 1976, the Cray-1 was state-of-the- art: it could process
160 million floating-point operations per second (flops) and Floating-point numbers had 8 megabytes
(MB) of memory. have decimal points in
them. Integers are simpler
Today, many personal computers can perform more than 10 to work with and require less times that number of
computing power.
floating-point operations in a second and have 100 times the amount of memory.
Meanwhile on the supercomputer front, the Cray XT5 Jaguar at Floating-point operation the Oak Ridge
National Laborat ory performed a sustained 1.4 petaflops in example: 2008 [source: Cray].
3.99233 x 45.210009 =
The prefix peta means 10 to the 15th power -- in other words, one quadrillion. That
Integer operation
means the Cray XT5 can process 8.75 million times more flops 10 x 13 = 130 than the Cray-1. It
only took a little over three decades to reach that milestone.
If you were to chart the evolution of the computer in terms of processing power, you would see that progress has been
exponential. The man who first made this famous observation is Gordon Moore, a co-founder of
the microprocessor company Intel. Computer scientists, electrical engineers, manufacturers and journalists
extrapolated Moore's Law from his original observation. In general, most people interpret Moore's Law to mean the
number of transistors on
a 1-inch (2.5 centimeter)
diameter of silicon
doubles every x number
of months.

The number of months

shifts as conditions in the
microprocessor market
change. Some people
say it takes 18 months
and others say 24. Some
interpret the law to be
about the doubling of
processing power, not
the number of
transistors. And the law
sometimes seems to be
more of a self-fulfilling
prophecy than an actual
law, principle or observation. To understand why, it's best to go back to the beginning.

A replica of the first transistor

AP Photo/Paul Sakuma
Semiconductors, Transistors and Integrated Circuits

The discovery of semiconductors, the invention of transistors and the creation of the
integrated circuit are what make Moore's Law -- and by extension modern electronics --
possible. Before the invention of the transistor, the most widely-used element in electronics
was the vacuum tube. Electrical engineers used vacuum tubes to amplify electrical signals.
But vacuum tubes had a tendency to break down and they generated a lot of heat, too.

Bell Laboratories began looking for an alternative to vacuum tubes to stabilize and strengthen the
growing national telephone network in the 1930s. In 1945, the lab concentrated on finding a way to
take advantage of semiconductors. A semiconductor is a material that can act as both
a conductor and an insulator. Conductors are materials that permit the flow of electrons -- they
conduct electricity. Insulators have an atomic structure that inhibits electron flow. Semiconductors
can do both.
 The control of the flow of electrons is what makes electronics work. Finding a way to harness the unique nature of
semiconductors became a high priority for Bell Labs. In 1947, John Bardeen and Walter Brattain built the first
working transistor. The transistor is a device designed to control electron flows -- it has a gate that, when
Gordon Moore
holds a vacuum closed, prevents electrons from flowing through the transistor. This basic idea is the foundation for the way
tube, the practically all electronics work.
precursor to
transistors.
Early transistors were huge compared to the transistors manufacturers produce today. The very first one was
half an inch (1.3 centimeters) tall. But once engineers learned how to build a working transistor, the race was
on to build them better and smaller. For the first few years, transistors existed only in scientific laboratories as engineers
improved the design.

In 1958, Jack Kilby made the next huge contribution to the world of electronics: the integr
ated circuit. Earlier electric circuits consisted of a series of individual components. Electrical
engineers would construct each piece and then attach them to a foundation called
a substrate. Kilby experimented with building a circuit out of a single piece of
semiconductor material and overlaying the metal parts necessary to connect the different
pieces of circuitry on top of it. The result was an integrated circuit.

Moore's Observation

The physicist Jean Hoerni developed the technique for creating planar transistors in 1959. By 1961, a company called
Fairchild Semiconductor produced the first planar integrated circuit. From that moment on, the technology advanced
rapidly. Physicists and engineers found new and more efficient ways to create integrated circuits. They refined the
processes they used to make components smaller and more compact. This meant they could fit more transistors on a
single semiconductor wafer than previous generations of the technology.

During this time, the director for research and development at Fairchild was Gordon Moore. Electronics magazine asked
Moore to predict what would happen over the next 10 years of development in the field of electronics. Moore wrote an
article with the snappy title "Cramming more components onto integrated circuits." The magazine published the article on
April 19, 1965.

Moore based his prediction upon the rapid development of the industry since the introduction of the integrated circuit. He
saw that as techniques improved and components on circuits shrank, the price for producing an individual component
dropped. Semiconductor companies had an incentive to refine their production techniques -- not only were the new
circuits more powerful, the individual elements were more cost efficient. As long as that relationship held true, Moore said
the trend would continue.

Moore also pointed out that over a 12-month period, the number of components on a one-inch (2.5-centimeter) diameter
semiconductor wafer doubled. Moore attributed this to two major trends: companies were finding ways to make smaller
components and they were getting better at arranging the components to conserve space on the wafer.

Moore extrapolated his prediction of the next 10 years by looking at this data. At the time of the article, Moore said that
the optimal number of components on a circuit was 50. He projected that by 1975, that number would be closer to
65,000. The prediction held -- by 1975, integrated circuits had nearly 65,000 components.

THE POWER OF CIRCUITS

One loose interpretation of Moore's law disregards the emphasis on how many components go onto a substrate. This version of the
law states that processing power for CPU chips doubles every 18 months. That's the neat thing about Moore's Law -- the meaning
shifts depending upon whom you ask.

Moore said the trend he predicted 10 years earlier would progress at that same rate for at least a few more years. But
Moore also said that he believed the semiconductor industry was approaching the limit for some techniques, such as
conserving space on a circuit. He called this factor "circuit cleverness." He believed that we'd reach a limit on how clever
we could arrange components -- eventually we'd have the optimal use of space. Once that factor is removed from the
equation, the rate of advancements must slow down. He said he believed after a few years components would double
only every 24 months.

While Moore's original observation focused on technological advances and the economics behind producing circuits,
many people reduce his observation to the simple statement we call Moore's Law. The most common version of Moore's
Law is that the number of transistors on a circuit doubles every 18 (or 24) months. Remarkably, this prediction has held
true -- today, Intel's Core i7 microprocessor has 731 million transistors, while its Xeon processor has 1.9 billion
transistors [source: Intel].

Cramming more components on an integrated circuit doesn't just mean devices are becoming more powerful -- it also
means they're getting smaller. The tiny components on compact integrated circuits power all sorts of portable electronic
devices. Even a small microprocessor chip today is as powerful as a full-sized chip was a few years ago. The advances
in circuit production make devices like smartphones and netbooks possible.

Moore's Law in Action

Why have Moore's observations and predictions held true over so many decades? Moore's Law isn't really a law at all --
in fact, there's no fundamental law of physics behind it. Moore's Law only holds true because of the actions of human
beings. But what keeps the cycle going even as the challenge to make more powerful circuits grows?

Much of the reason is psychological and driven by the market. Companies that make integrated circuits are competing
against each other and everyone knows about Moore's Law. That means every corporate executive has this in mind: if
our company doesn't double the power of our circuits in 18 months, another company will beat us to it.

Because companies don't want to give an edge to competitors, they pour a lot of money into research and development
(R&D). These R&D divisions work to develop new techniques to create smaller components and arrange them in such a
way that maximizes their performance. It costs a lot of money to keep up the cycle of research, but this cost is balanced
against the threat of competitors gaining a foothold and dominating the market.

Another factor is the simple desire to overcome a challenge. Many people have predicted the end of Moore's Law over
the years. Some people thought it would come to an end during the 1980s. Others said the same thing in the mid '90s. It
seemed like engineers would eventually bump up against a barrier that would be fundamentally impossible to breach.
But engineers somehow manage to find a solution each time, keeping Moore's Law alive.

Consumers also drive Moore's Law. The rapid development of electronics has created a sense of expectation among
consumers. Every year, faster and more advanced electronics hit the market. From the consumer's point of view, there's
no reason not to expect something better next year.

Moore's Law and the Nanoscale

Today, transistors on integrated circuits have reached a size so small that it would take more than 2,000 of them stacked
next to each other to equal the thickness of a human hair. The transistors on Intel's latest chips are only 45 nanometers
wide -- the average human hair is about 100,000 nanometers thick [source: National Nanotechnology Initiative].

Creating such narrow transistors is an amazing achievement. Visible light has wavelengths in the range of 400-700
nanometers. Conventional light microscope lenses can only focus on objects half the size of the wavelength of visible
light or larger. You have to rely on special equipment like scanning electron microscopes to create an image of
something on so small a scale.
A nanometer is 1
One thing you must consider when dealing with such small devices is that as you approach
billionth of a meter. The
the nanoscale, you leave behind the world of classical physics and enter the realm of diameter of a helium
quantum mechanics. The rules of physics in the quantum world are very different from the atom, for example, is
way things work on the macro scale. For example, quantum particles like electrons can pass about 0.1 nm, and that
of a ribosome is about
through extremely thin walls even if they don't have the kinetic energy necessary to break 20 nm.
through the barriers. Quantum physicists call this phenomenon quantum tunneling.

Because electronics depend upon controlling the flow of electrons to work, issues like quantum tunneling create serious
problems. These problems force electrical engineers to re-evaluate the way they design circuits. In some cases, shifting
to different materials solves the issue. In others, finding a completely new way to build circuits might work.

There's even the possibility that someone will come up with a revolutionary idea that makes the transistor and integrated
circuit obsolete. While that may sound far-fetched, the fact remains that despite numerous tech pundits and engineers
pronouncing the end of Moore's Law, circuit manufacturers are still finding ways to keep it going. As it turns out, the
challenges may not be quite as impossible as some believe.

The Future of Moore's Law

Even Gordon Moore has questioned how long the cycle of innovation and production can keep up the frenzied pace of
the last four decades. He has also expressed amazement at the way companies like Intel find new ways to work around
what initially seemed like an insurmountable problem. Will there ever be an end to Moore's Law?

The answer is yes, but it's difficult to pin down when that might happen. For one, we could hit a technical barrier that
prevents engineers from finding a way to make smaller components. But even if we don't encounter a technical barrier,
economics could come into the equation. If it's not economically feasible to produce circuits with smaller transistors there
may be no reason to pursue further development. Or we could bump up against the fundamental laws of physics -- like
the speed of light, for instance.

The problem with predicting a specific date when one or more of these barriers will stop progress is that we have to base
it on what we know today. But every day engineers are learning new ways to design, build and produce circuits. What we
know tomorrow may make the things that seem impossible today completely achievable.

Is Moore's Law even relevant today? The era of the personal computer has been dominated by a sense that the
consumer needs the latest and greatest machine on the market. But today, some people are questioning that philosophy.
Part of that is due to changes in consumer behavior -- many computer owners use their computers for simple tasks like
browsing the Web or sending e-mail. These applications don't put a heavy demand on the computer's hardware.
Another reason powerful PCs aren't as necessary is the rise in popularity of cloud computing. Cloud computing shifts
the burden of processing and storing data to a network of computers. Users can access applications and information
using the Internet, so they don't necessarily need a powerful machine of their own to take advantage of cloud computing.

As a result, devices like smartphones and netbooks are becoming more popular. These devices don't have the raw
processing power of the latest desktop and laptop computers. But they still allow users to access the applications and
data they need.

Lessons Learned from Moore's Law

If consumers continue to purchase devices like smartphones and netbooks, microprocessor manufacturers will have less
of an incentive to meet the expectations of Moore's Law. If there's no market for ultra-powerful processors, then we've hit
the economic barrier that could bring an end to the cycle.

That said, some facilities may still push the limits of integrated circuit production. While the average consumer may not
see the value in a powerful PC, research facilities still rely on the fastest processors in production. More
powerful microprocessors can aid in everything from weather prediction to cosmological studies.

One lesson we can draw from Moore's Law and the semiconductor industry is that pure research can yield beneficial
results for society. The engineers at Bell Laboratories had no guarantee that their experimental work with the earliest
transistor models would yield positive results. But their research and hard work spawned an industry that changed the
way we live. It's an example of how scientific research can have a dramatic impact on our lives even when there's no
obvious or immediate benefit.

Perhaps the most important lesson we can take from Moore's Law is that we shouldn’t be too quick to say something is
impossible. Henry L. Ellsworth, the commissioner of the U.S. Patent Office in 1843, once
said "the advancement of the arts, from year to year, taxes our credulity and seems to
presage the arrival of that period when human improvement must end" [source: Sass].
Ellsworth was pointing out that the rate of human inventiveness and innovation was so
impressive that it was hard to believe. He was not, as some have implied, suggesting that
everything that could be invented had been invented already. And as a matter of fact, the
rate of innovation has only increased since then.

While we know a great deal about building electronics, there may yet be very much we do
not know. Moore's Law helps as a motivational device for inventive engineers. They don't
want to disappoint Gordon Moore, even if it means they have to find unique solutions to
seemingly impossible problems. Henry L. Ellsworth – U.S.
Patten Office Commissioner
1843
 A Good Paying Job with Exponential Growth
Name: Kaiya Little Class Period: 1

You have been offered two different jobs. Which would you accept? Do the math below and then circle your choice.

1) Your boss agrees to pay you $50 per hour. You can work 8 hours a day for one month.
2) Your boss agrees to pay you 1 cent per day but will double your pay every day that you work.

Job 1 : $50 per hour

1. Write a math expression to calculate the number of hours worked in 30 days. 8x30x50
2. How much pay would you earn in 30 days at $50 per hour 12000

Job 2 : $0.01 for the first day. Double salary each day.
Day 1 0.01 Day 11 10.24 Day 21 10421.76
Day 2 0.02 Day 12 20.48 Da y22 20843.52
Day 3 0.04 Day 13 40.96 Day 22 41687.04
Day 4 0.08 Day 14 81.92 Da y23 83374.08
Day 5 0.16 Day 15 162.84 Day 23 166748.16
Day 6 0.32 Day 16 325.68 Da y24 333496.32
Day 7 0.64 Day 17 651.36 Day 24 666992.64
Day 8 1.28 Day 18 1302.72 Da y25 1333985.28
Day 9 2.56 Day 19 2605.44 Day 25 2667970.56
Day 10 5.12 Day 20 5210.88 Da y26 5335941.12

1. Since early days of the pay schedule make little difference in the total amount, add only the last 10 days. You
may round numbers to make the math easier. What is the estimated total pay? 10661460.48
2. Draw a line chart showing how much money you would earn each day. Work backwards from day 30 to day 1.

Pay

1 2 3 4 5 6 7 8 9 10 11 12 13 14 Day
15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Number