Gordon Moore Just Died, Will His “Law” Long Outlive Him?


New member
Co-founder of processor giant Intel, American Gordon Moore, has died at the age of 94. Having long since left the control of the semiconductor giant, he nevertheless left behind a law in history that bears his name. And what remains at the heart of the industry’s concerns: the continued increase in the power of our computers.

Intel is in mourning. Gordon Moore, co-founder and emblematic leader (from 1975 to 1987) of the processor giant, died March 24 at age 94, Intel and the Gordon & Betty Moore Foundation announced. If you follow processor news, you may be familiar with Moore’s “Law”. Indeed, it was Gordon Moore who – by observing technological developments – theorized about the frequency of the doubling of the number of transistors and thus the performance of microprocessors. This law, last revised in 1975, was initially drawn up for a period of ten years and provided that processors’ assets would double every two years. And although he is harassed by technical limits and physics itself, this “law” is still valid.

A photo from 1978. Pictured, from left, are Andy Grove, Robert Noyce, and Gordon Moore. Moore and Noyce co-founded Intel Corporation in July 1968. Grove was instrumental in the company’s success. Moore has served as President Executive Vice President, Managing Director and board member. (Credit: Intel Corporation)

Gordon Moore is one of the pioneers of the semiconductor world. Before founding Intel in 1968, he worked for one of the inventors of transistors, William Shockley. As for Fairchild Semiconductor, a company that ended up like Philips in the Netherlands or Thomson in France by spreading its talents to the rest of the country, Intel is one of Fairchild Semiconductor’s direct heirs. The success of the company he helped launch, as well as the quality of his predictions, has made Gordon Moore a giant in the world of ICs – for integrated circuits (integrated circuits). This is also a myth for Intel. Thus, in 2022, the group’s management gave its name to the new campus of its plants in Oregon in the United States.

In this undated photo, Gordon Moore works at Intel.  Intel and the Gordon and Betty Moore Foundation announced that the company's co-founder passed away on March 24, 2023.  The author of Moore's Law was 94 years old.  (Credit: Intel Corporation)

In this undated photo, Gordon Moore works at Intel. Intel and the Gordon and Betty Moore Foundation announced that the company’s co-founder passed away on March 24, 2023. The author of Moore’s Law was 94 years old. (Credit: Intel Corporation)
A law that died a thousand times

The end or “death” of “Moore’s Law” has been announced every year for more than two decades. Last year it was the CEO of Nvidia who promised his end during a press conference. However, 2023 will be the year when mass production of 3nm chips will start and performance will continue to increase at an ever-increasing pace, albeit at a slower rate than in the past. No more tens of percent automatically gained by the reduction of circuits, today engineers use all possible areas of progress: in addition to engraving, there is the complexity of architectures, the heterogeneity of logic elements (SoC, read below), chip stacking (potato chips), or the software.

Regular evolution of the number of transistors, Moore's law still holds.

Should Moore’s law be changed? A change to every three years might be appropriate. But the reality is that overall chip capabilities are much more difficult to measure than they were in the past. While Moore’s law was about the power of CPUs, the chip that powers your phone is a veritable Frankenstein’s monster, made up of several “pieces”. SoCs called for system on a chip (system on a single chip), these super components simultaneously integrate CPU, GPU, ISP or NPU (neural processor). And not only the processors are heterogeneous in type of components, but they are also in the fineness of the engraving, from now on more important levers for the improvement of performance. And this reduction in circuit size that allows the increase in transistor density per mm² is, if not threatened, at least increasingly complicated to achieve.
The Challenges of the Angström Era

When Gordon Moore founded Intel, circuit sizes were measured in tens of micrometers. When he left the company’s management in 1987, the industry had just entered the nanometric era, with chips engraved in 0.8 microns, or 800 nm. Mr. Moore has just passed away at a time when the industry should be supplying a mass of chips engraved 266 times finer in 3nm. An incredible achievement that, however, should not obscure the “walls” that await the world of semiconductors.

Because the transition to the “Angström era” as Intel calls it, will not be as easy as in previous decades. If micrometer (10-6) we then moved on to nanometers (10-9), the next unit is more modest – Angstrom, a non-IS unit representing 10-10. One way to avoid commas: after 2 nm, Intel delivers 18 Angstroms (18A), which is equivalent to 1.8 nm.

Read also: Intel accelerates its “Angstrom-era” production plans and is ahead of schedule (April 2022)​

Why change units? Because the next steps are real leaps into the blue. Going from 3nm to 2nm isn’t a simple subtraction from “1”: it represents a 33% reduction in circuit size! For scientists, engineers and other researchers, a first major opponent: physics and the limits of human knowledge. The entire industry is struggling to improve manufacturing processes, whether it’s developing new machines, new software or finding new materials. If physics is the ultimate engineering enemy, there’s another one that’s deeply human: economics.

Will Moore’s Law Collapse on a Rock?

Chip development costs

Chip development costs

After Moore’s law took root in people’s minds, another law emerged: that of production costs. So said Arthur Rock, an early investor in semiconductors who had bet on companies like Intel, of course, but also Apple or Teledyne. Lack of luck for us, his law is formulated as follows: The cost of semiconductor manufacturing plants doubles every four years “. And the man, who is now approaching a hundred years, hits the nail on the head: factory costs are skyrocketing. Intel’s future German location in Magdeburg should therefore cost around 30 billion dollars.

Read also: Why the race to miniaturize our chips will slow down in the coming months (January 2023)
While current inflation is clearly driving up prices, it is mainly the incredible complexity and technical nature of the processes that are to blame. An analyst firm recently estimated that the cost of developing a 5nm chip is about half a billion dollars… before production even starts. The production in question is now, for the most advanced chips, based on state-of-the-art machines of which the essential scan (Or stepper) EUV of the Dutch ASML. The only machine in the world that can etch circuits with extreme ultraviolet radiation – so all chips below 5 nm. These precious machines, sources of all greed, trade for about $180 million each… and it takes several of them to run TSMC’s and Samsung’s mega-factories, which release millions of chips each month.

If these machines are so expensive it is because of the nature and precision of the instruments – the lenses of the optics that focus the rays are jewels, the mirrors are polished to provide a flat surface at more or less 20 picometers! But beyond their purchase price, ASML scanners are energy consumption monsters: by the kiloWatts, then hundreds of kiloWatts, of previous EUV scanners, the EUV era increases consumption to 1.5 megaWatts per machine!

Gordon Moore was the co-founder of Intel Corporation and the author of Moore's Law.  He co-founded Intel Corporation in July 1968. (Credit: Intel Corporation)

Gordon Moore was the co-founder of Intel Corporation and the author of Moore’s Law. He co-founded Intel Corporation in July 1968. (Credit: Intel Corporation)

And while ASML and its partners are working on the next generations of machines, current CEO Martin van den Brink explained to bits-chips.nl last September that “ technically it is possible [mais] or the cost of hyper-NA scanners (currently under development, editor’s note) increases similar to what we saw for high-NA scanners (currently on the cutting edge, editor’s note), it just becomes economically unfeasible”, explains the one who has overseen the success of today’s high-NA EUV scanners. In other words, the survival of Moore’s famous law is in his hands. But Gordon Moore can no longer die: he has made history.

Source :