Moore’s Law Isn’t Dead, But It Could Become Useless

In July, when Intel announced it was delaying mass production of processors using its next-generation 10nm manufacturing process, many thought it was anything but confirmation that Moore’s Law – an empirical observation of the number of components that could be built on an integrated circuit and their corresponding cost – is dead. In 1965, about twenty years after the invention of transistors and five years after the invention of the integrated circuit, chemist and scholar Gordon Moore made a single logarithmic plot on a sheet of paper predicting that with the advent of microelectronics , the computing power would increase significantly while also becoming much cheaper over time. The very simplified version was that computing power would double every two years as chips became denser.

The observation, which has remained true for years and decades, is beginning to show unfortunate inconsistencies. But is it really dead?

“Integrated circuits will lead to marvels such as personal computers or at least mainframe-connected terminals, automatic controls for automobiles, and personal portable communications equipment,” Moore wrote in a now famous article for Electronics Magazine. . This observation has shaped our lives, touching almost every aspect of it. What was equally fascinating about Moore’s Law was that it followed no scientific principles. It was something else entirely.

(Also see: Tech 101: Explaining the SoC)

“The whole nature of this curve as a kind of law is a very interesting phenomenon because it’s not a phenomenon like a law of physics. It’s really a phenomenon about what people are willing to believe. “, said Dr. Carver Mead, Silicon device visionary at the California Institute of Technology in a video produced by Intel. “And what that did was it gave people the confidence to take the next step, and after a while those things become fulfilling because people believe they can do it, so they often do what’s necessary to make it happen, and that’s what actually happened with Moore’s Law.

But even Moore didn’t think his observation would remain accurate for such a long time. The extrapolation made by Moore in 1965 maintained that the complexity of integrated circuits would increase by a thousand times in ten years, compared to what it was at the time. But in 1975, Moore presented another paper in which he noted that the industry had already achieved that goal, leaving very little room to pack things more densely, at least as quickly. Moore then noted that in the future, chipmakers would need two years to make processors twice as complex.

The first stumbleBut something interesting happened. Instead of the 24-month timeframe, chipmakers managed to double complexity every 18 months, and continued to do so for more than two decades before changing the timeframe to two years. This revised schedule remained largely unchanged until recently, when chipmakers finally began to struggle to move forward at the required pace. Intel’s transition from 22nm to 14nm, for example, was delayed after the chipmaker ran into difficulties with lithography. While Intel then assured the world it would pick up the pace and return to a 24-month cycle, it has now noted that a transition to 10nm would take around 30 months.

Additionally, in September, Intel announced that it had committed $50 million and engineering resources to a 10-year partnership with a Dutch university to further research into quantum computing. The company’s leap to an entirely different model indicates that it understands that it might, at some point in the future, hit a wall when it comes to packaging transistors more densely. Quantum computing is totally different from transistor-based computing, on which Moore’s law is based. It uses quantum entanglement to enable a combination of binary inputs using qubits, the quantum bits of information. Intel doesn’t want to become useless even if the transistor itself does.

(Also see: Tech 101: What is a processor? Part 1 – Logic Units, Instruction Sets, Microarchitectures)

Looking at the rest of the industry roadmaps, for example, it’s clear that their development process still follows the age-old Moore’s Law. Taiwan Semiconductor Manufacturing Corp., which, as its name suggests, manufactures semiconductors for many companies that do not have their own manufacturing facilities, began volume production of 16nm processors in the second quarter of this year and is on track to offer 10nm production line in the first quarter of 2017. The company also announced that it will start manufacturing 7nm processors in the same year.

Samsung is also shipping 14nm FinFET chips starting this year and plans to begin mass production of 10nm chips by the end of 2016. Earlier this year, IBM unveiled a new ultra-dense chip design, which, as he claimed, is four times more powerful and twice as advanced as anything available on the market today. These developments suggest that Moore’s Law is not in such bad shape after all.

Raj Sabhlok, president of Zoho Corp., has another theory. He notes that while the pace of desktop hardware has slowed down over the past two years, something interesting has taken over. “The price of software applications has fallen, while functionality and quality have increased exponentially,” he wrote in Forbes.

What happens afterwards?Many in the industry wonder how long it will be possible to continue to follow Moore’s Law. The minimum cost per transistor has steadily increased since the advent of 28 nm chips on the market. The cost of newer and more sophisticated photolithography equipment needed to manufacture small integrated circuits is also increasing. What’s even more troubling is that the benefits of continuing to shrink transistors don’t seem to be as great as many would like. After a certain point, it may not be worth continuing to get smaller.

(Also see: Tech 101: What is a processor? Part 2 – 64-bit, number of cores and clock speeds)

At the same time, consumers’ growing reliance on cloud computing gives chipmakers fewer reasons than ever to make certain types of processors more complex.

As long as a web browser allows us to do what we need, a gigantic processor may not be necessary at all. In the future, we may be able to offload most of the tasks that require processing on our devices to a stack of servers placed in a remote data center. Cloud-based apps and games are big business now, with established companies like Nvidia with Grid and startups like SWYO jumping on board.

Of course, the network infrastructure will have to improve considerably, but over time, we may no longer need a high-end device. What we would need is something capable of connecting to the internet with just enough processing power to handle the browser software.

Going back to today, many people are quite happy with the computers and smartphones they have. Companies hope that new features and experiences such as 3D, virtual reality, natural language interaction and even artificial intelligence will drive demand for new computers. If all you care about is email and word processing, you’re fine with what you already have (but you’ll be glad new high-end advancements like these usually drive down the prices of basic devices).

Moore’s Law has obviously slowed down, but if Intel and the industry as a whole are committed to overcoming new challenges, its essence could remain relevant for many years to come.