100 Years of Moore’s Law
We’ve seen a flurry of articles on Moore’s Law recently in celebration of its 50-year anniversary this month. Most are quite good, especially Larry Downes’ piece in the Washington Post. The only one I didn’t like was in Pando Daily because it offered the non sequitur “gee, if Moore’s Law stops we could really focus on writing better apps.” The people who design chips don’t write apps, so there’s really no connection of that kind.
There seems to be some misunderstanding about what Moore’s Law is, and how long it remains in effect. Gordon Moore himself projected it would last about a decade, since it’s such an unusual phenomenon; Carver Mead predicted it would last 40 years or so based on laws of physics.
To put all of this in perspective, bear in mind that integrated circuits (computer chips) are means of controlling electrons in a solid, and there are lots of ways to do this. Moore’s Law – the prediction that integrated circuits would become twice as dense every two years, more or less – means we can control electrons in a solid much better than we used to.
It’s true because humans make it true. As Mead (the man who coined the term “Moore’s Law”) said ten years ago: “It became an almost religious faith in human ingenuity and a belief in the future.”
Gordon Moore’s prediction of exponential improvement was based on experience: He had observed a doubling of integrated circuit capacity every couple of years (18 months in some accounts) and simply expected that rate of improvement to continue for a decade.
This came about for two reasons: improvements in the equipment that make ICs and the desire to leapfrog competition. ICs are a new technology that depends on a very complicated series of steps that include circuit design, silicon crystal creation, photolithography, and chemistry.
Making an integrated circuit is a lot like old-school photography before digital: You grow a pure silicon crystal in a lab, slice it into wafers, paint a pattern on it with chemicals, and etch circuits on the wafer by exposing the chemical pattern to light. The chemicals left behind form transistors, resistors, and capacitors, the elements of logic switching circuits.
After chemicals are deposited, excess is washed off, more chemicals are applied which are exposed to light again, and more circuits are created. This continues until you have billion transistors on a single silicon wafer. In the end, you enclose the wafer in plastic or ceramic and somebody puts it on a circuit board.
This process creates an electrical circuit because the silicon is an insulator and the chemicals are conductors with particular properties that allow actual circuits to be created. Improvements come about from the ability to perform any of the steps with more precision, and they continue to the physical limits of the materials. This is to say that circuits can’t get any smaller than atoms (as far as we know) and silicon only has so many atoms per cubic millimeter. An insulator stops being an insulator if it’s too thin.
We’re approaching the limits of miniaturization on silicon crystals, so some are predicting and end of Moore’s Law as we know it. That’s not necessarily an end to improvements in integrated circuits, however: Other materials such as gallium nitride will allow circuits to get smaller, but at the moment it’s more expensive than silicon, which is after all the most plentiful element on earth.
Assuming we do come to the end of miniaturization some day, even that doesn’t mean we’re at the end of the road for tech innovation. The ingenuity behind integrated circuits is motivated by the pure capitalist desire to make money by building better products, and that force is stronger than silicon. There are also many ways to improve chips without making circuits smaller.
For example, companies can make money by simply organizing an attractive set of circuits on a chip, as they do when making a so-called application-specific integrated circuit, or ASIC. When a company makes a single chip that does everything a low-end smartphone needs to do, there are plenty of buyers.
That chip can be improved by making it more power-efficient, more rugged, or faster. These ends can be accomplished with clever logic design. And companies can figure out how to make chips cheaper — another sure winner.
The staggering feature of Moore’s Law is the exponential improvement. This doesn’t happen anywhere else in the economy. For example:
- Average yields of corn have increased by two percent a year since 1950;
- The generation of electricity from steam improved by 1.5 percent annually in the 20th century;
- Outdoor lighting efficiency has improved by 3.1 percent annually over the past 135 years;
- The speed of intercontinental travel improved by 5.6 percent per year from the 1900 ocean liner to the 1958 Boeing 707, but has been flat since;
- Between 1973 and 2014, passenger car fuel efficiency has improved by 2.5 percent annually; and
- The energy cost of steel declined by 1.7 percent per year between 1950 and 2010.
The only time we see exponential growth outside of Information Technology industries is when traditional industries adopt IT methods. This is happening in electric power, as solar panels built on methods pioneered in integrated circuit manufacturing transform the way electricity is generated. There’s actually such a rapid take-up of solar panels in some areas that electric utilities are freaking out and trying to slow people down.
The most important feature of Moore’s Law is the expectation of continual improvement. Because everyone in the integrated circuit business expects competitors to produce exponential improvements, everyone strives to make them. And everyone who makes chips is able to do better because the companies that make the machines that make chips uses chip technology to make better machines. Et cetera.
We’re about to see, as firms like Apple and Google eye the success Tesla has had with high-end electric cars. Since the modern car is more like a computer than a traditional car is, this will be exciting to see.
Happy birthday, Moore’s Law. May you live to 100!
Do you think Moore’s Law is infinitely sustainable? Or will the idealistic theory bump up against the reality of physical limits? Sound off in the forum!