Is Moore’s Law Broken?
For 50 years, Moore’s Law has driven computer speed and power like a fixed law of nature. But now it seems its decades long reign is coming to an end.
Back in 1965, computers still took up entire rooms and cost as much as a house. But Gordon E. Moore was able to look at all the progress they had made at the time and project it into the future. It had been less than 10 years since he and seven other colleagues had founded their semiconductor business deep in the Santa Clara Valley. There, they would design and fabricate transistors and integrated circuits — without which, almost all modern electronics would not be possible.
Moore observed that microchip complexity — the number of transistors packed into an integrated circuit — had doubled every year since 1959, and he saw no reason to think that this growth wouldn’t continue unabated for at least another decade.
This is Moore’s Law.
And that simple projection would dictate computer design and performance for decades to come.
The phone or computer you are reading this article with has hundreds of millions of transistors. Probably upwards of a billion or more.
There have been a few iterations of Moore’s Law over the years.
In 1975, ten years after Moore’s Law debuted, Moore revised it to say transistor density would double every two years — instead of every year.
Then David House, an executive at Intel, adopted it and revised it to mean computer performance instead of transistor density. And in House’s version of Moore’s Law, computer performance would double every 18 months.
And this is really what we think of when we talk about Moore’s Law. Faster, better, smaller, computers.
But Moore’s law isn’t so much a law like Newton’s laws as it is a guideline.
It’s a benchmark that every computer manufacturer and manufacturer of computer components could aim for.
Moore’s Law was a goal to be met.
So in the course of those five decades, it’s hard to say whether Moore’s prediction was spot on, or if the computer industry merely conformed to Moore’s projection.
Each year they could look to Moore’s Law, and everyone in Silicon Valley knew where they needed to be. And this extended to the entire world. Moore’s Law facilitated an unprecedented amount of cooperation across an industry populated by thousands of different companies, building a wide variety of products — often in competition with each other.
Never before had an industry worked in concert with such precision, united under a single edict — Moore’s Law.
So in that sense, the electronics industry turned Moore’s Law into a self-fulfilling prophecy.
But just because the industry drove Moore’s Law as much as Moore’s Law drove the industry didn’t mean it could go on indefinitely.
Moore’s Law is exponential.
But the technology itself has limitations. And we’ve run up against some of those limitations in recent years.
Although we’ve come to expect increased performance with increased transistor density, Moore’s Law never meant for the two terms to be interchangeable. In recent years the computer industry has witnessed a decoupling between these two benchmarks — with increased transistor density leading to only minimal performance increases.
Evidence of the slowing progress of Moore’s Law can be found at the company Moore himself founded — Intel.
Intel has delayed future transistor improvements and increased the time between upgrades. The company has even suggested that silicon-based chips will stop shrinking altogether within 5 years.
We’ve come to expect computers to get smaller and lighter every year. This is a direct result of making smaller microchips ever more densely packed with transistors — which is what Moore’s Law decrees.
But there is a limit on how small we can go.
One impediment is heat. To make your computer faster, you have to make the electrons in the circuits run faster. But the faster you move your electrons around the circuit, the more heat you generate.
And heat is hard to get rid of.
But if you don’t want to get a third degree burns from your smartphone, you have to put a speed limit on your processor. Companies have been doing just that for over a decade. In fact the maximum clock speed in a computer plateaued in 2004 and has remained unchanged ever since.
To make up for the loss in processor speed, companies have found a workaround. They increase the number of processors.
Dual core, quad core, even eight cores are common in today’s computers and it’s a direct result of putting the brakes on processor speed.
Multiple separate processor cores act pretty much like a single, super fast core except any problem the computer needs to compute has to be divvied up amongst all the processors. This means, if there are eight cores, there has to be eight algorithms working in parallel on a single problem. This can be tricky — to say the least — and even impossible at times.
There’s also a limit on how close the electrons can be spaced. Too close and the efficiency of the chip degrades. Thus, there’s a limit to how dense the transistors can be on the circuit. And if the transistors are built too small, they begin exhibiting quantum behavior. Then the electrons become unpredictable and the chips themselves become unreliable.
Then there’s the ultimate limit. You can’t build transistors smaller than an atom because they are themselves made up of atoms. They can’t be smaller than their constituent building blocks.
We haven’t run into this limit quite yet but it’s coming in the next decade or so.
But we might just run out of money before that.
Moore’s 2nd law states that the cost to manufacture exponentially more complex microchips increases exponentially. So someday it will be prohibitively expensive to build a smaller, denser microchip.
But this is more than just a technological problem.
Economic growth is tied to productivity. And over the last few decades, productivity has been fueled by computer power. Companies and research labs who depend on faster, better, smaller computers for their advancement could be hindered by limited processor speeds.
And if this year’s newest model is only marginally better than last year’s, consumers will be more likely to stick with their old computers and skip purchasing the new model.
As Moore’s Law breaks down, so does the engine that has been fueling much of our economy.
So if we are to overcome the limitations of Moore’s Law, we might need to abandon silicon for another, better material.
Some researchers are exploring exotic forms of carbon, like graphene, which are only one atom thick. Or they are attempting to harness electron spin to process information more efficiently. Or there’s the promise of quantum computing which would utilize the weird behavior at the smallest realms of universe to make processors that can handle multiple computations at the same time.
But whether it’s whole new materials or whole new physics, the fact is most of these proposals have barely made it out of the lab and there’s no way of knowing which one will be the winner. For the first time in 50 years, the future of computers is uncertain.
And there’s no Gordon Moore-like prophet to show us the way.
But until we find that new material or technology, we are stuck with silicon. To make up for a lack in processor progress, computer and electronics manufacturers have to become more clever.
And that means relying on gimmicks to get us to buy new products.
More apps, flashier designs, and new add-ons all take the place of more tangible improvements. While these extras can enhance the user experience or make a device easier or more enjoyable to use, they don’t necessarily improve the computer’s performance by much.
The entire computer industry has been based on the premise that every new model is leap forward in technology. And that premise has been based on Moore’s Law.
With the end of Moore’s Law in sight, maintaining this premise becomes more and more a technological sleight of hand — with no real improvement happening at all.
Before Silicon Valley filled up with scientists and startups, the main industry was agriculture.
The valley was home to farmers and fruit trees. These orchards eventually gave way to microchip plants and silicon became the bumper crop.
Now silicon is on the way out. What its successor might be cannot be known for sure. But if the unbridled, sometimes reckless optimism of Moore’s Law has taught us anything, it’s that the material itself doesn’t matter. Whether it’s fruit or silicon or some cutting-edge technology, the deciding factor will be whether we can work together, united by a common purpose.
For 50 years we’ve had Moore’s Law to guide us and to keep us on track. It allowed thousands of disparate companies, all competing with one another for market share, to cooperate and progress in unison. To work as a single unit for a common goal. And this unprecedented unity in purpose gave us one the greatest technological booms in all of human history.
Because Moore’s Law is ultimately about optimism.
It was about looking toward the future and not just hoping things could be better, but knowing they had to be better. It was the rightful charge of the computer industry to accomplish the mighty commandment set forth by Moore’s Law.
And the only aspect of Moore’s Law that can’t be broken, that we can’t revise, is its fearless pursuit of progress.
Video version here: https://www.youtube.com/watch?v=PA2QHDQBh8A