Professor Li mentioned that Moore’s Law is continuing and that there is no sign of stopping. However, this depends on which angle you are looking at things. Considering the generally accepted definition of Moore’s Law, it is stated that in 1965, Gordon Moore proposed that the number of transistors on a silicon chip would double every year. Because of this, we have experienced an exponential growth of computing power that has made much of the technology of today possible.
Change from the focus on hardware to software
For over the past 50 years, the focus has been on the physical side by reducing the size of transistors. Currently, the industry has accomplished making the transistors as small as 10 nanometer, which is so small that new complications have arisen. The research and development departments are dealing with electrical leakage, increased generated heat, and expanded cost of cooling. These problems are the main causes of contributing to the slowing rate of growth in processor power. It is estimated by several experts that Moore’s Law will end around 2025 because transistors will have reached their physical limits. Will this mean that technology advancement is all gloom and doom?
Not necessarily. Because it has been relatively easy to focus on increasing processor power by expanding the number of transistors on a single chip, there never really was any incentive to optimize performance out of a single chip. As the human species naturally seeks continued advancement, I believe that pushing more performance out of the same single chip will become much higher of a priority. This could translate to better algorithms, to optimize both efficiency and speed. We simply do not have a choice. How much and for how long this switch will lead to improved performance remains to be seen.
How do you think Moore’s Law will be continued? What are your predictions regarding the impact of switching to a software focus for Moore’s Law?
Let me know in the comments below!
References:
Rotman, D. (2020). We’re not prepared for the end of Moore’s Law. Retrieved on 10 September 2020 on https://www.technologyreview.com/2020/02/24/905789/were-not-prepared-for-the-end-of-moores-law/
Loeffler, J. (2018). No more transistors: The end of Moore’s Law. Retrieved on 10 September 2020 on https://interestingengineering.com/no-more-transistors-the-end-of-moores-law
Moore, S., K. (2019). Another step toward the end of Moore’s Law. Retrieved on 11 September 2020 on https://spectrum.ieee.org/semiconductors/devices/another-step-toward-the-end-of-moores-law
Bailey, B. (2018). The impact of Moore’s Law ending. Retrieved on 11 September 2020 on https://semiengineering.com/the-impact-of-moores-law-ending/
Tardi, C. (2020). Moore’s Law. Retrieved on 12 September 2020 on https://www.investopedia.com/terms/m/mooreslaw.asp
Dear Dennis,
Thank you for your nice blog posts. I agree that someday there will be an end for Moore’s law, but it’s difficult to pin down when that actually might happen. The problem with predicting a specific date is that we have to base it on what we know today. But everyday 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.
An interesting question to raise is; What happens when Moore’s Law inevitably ends?
I think technology companies should research new successor technology investments now to find out where to invest in next. Because one prediction is pretty much certain to come true: we’re always going to want more computing power.
Kind regards,
Danielle van Helden
Hi Dennis,
thanks a lot for bringing up this interesting question on Moore’s Law. I totally agree with you that, once transistors will have reached their physical limits, a new focus on algorithm efficiency will be necessary to accomplish further technological advancements.
But what if transistors may become obsolete before their physical limits are reached. The new disruptive technology of Quantum Computers is on the horizon and it promises unprecedented computing power. The urgent need to focus on software efficiency and optimize performance out of a single chip would vanish.
Hartmut Neven, director of the Quantum Artificial Intelligence lab said at the Google Quantum Spring Symposium that quantum computers are gaining computational power relative to classical ones at a “doubly exponential” rate. That means that we don’t recognize any progress for quite some time. But suddenly, we find ourselves in a different world of computing. IBM announced in 2019 that they have a quantum computer ready for commercial use by organizations (not yet individuals).
So maybe the mentioned explosion of computer power is not as far away as it may seem?
Perez-Delgado, C. (2019). IBM launches commercial quantum computing – we’re not ready for what comes next. Retrieved September 25, 2020, from https://theconversation.com/ibm-launches-commercial-quantum-computing-were-not-ready-for-what-comes-next-110331
Harnett, K. (2019). A New Law to Describe Quantum Computing’s Rise? Retrieved September 25, 2020, from https://www.quantamagazine.org/does-nevens-law-describe-quantum-computings-rise-20190618/
Rossi, A. and Gonzalez-Zalba, M.F. (2019). We Need to Replace Moore’s Law to Make Way For Quantum Computers, But What’s Next? Retrieved September 25, 2020, from https://www.sciencealert.com/why-it-might-be-too-soon-for-a-moore-s-law-for-quantum-computers
Hi Dennis,
Thank you for sharing this blog post with us. Your questions in the end really got me thinking. I even did some research on the subject after reading your post which I’d like to share with you together with some personal thoughts.
I totally agree that Moore’s Law in the sense that we all know it is coming to an end in the near future. Unless there is some major technological breakthrough, physical constraints will slow down – or even stop – the ‘who can fit the most transistors in one chip’ challenge. By the way, did you know that Gordon Moore revised his own law 10 years after his initial statement? He then said that the number of transistors would double every two years, instead of one (Rotman, 2020).
You ask how Moore’s Law would be continued. To be honest, probably only a few people know. Nevertheless, I advocate your angle on the case. I foresee the performance-focus you mention to grow in popularity and in practice. The implications thereof will become more clear when the time is there. However, I think that one implication is that it will give rise to new world-leading companies who excel in this novel area.
Next to the aforementioned, I expect a more radical change in the sense that silicon computer chips will slowly disappear over time. There are various contenders to replace the chips, such as quantum computing, graphene and carbon nanotubes and nanomagnetic logic (James, 2018). Maybe, some day, another ‘Moore’ will predict great improvements for one of these technologies or for something that hasn’t even been developed yet. And maybe, just maybe, we’ve underestimated Gordon Moore, and Moore’s Law will continue to be true. This certainly wouldn’t be the first time people questioned the strength of his law and found themselves at the wrong side of the coin 🙂
Thanks again for getting me think about this topic! If you are interested in some reading material, here are the sources I read and used:
– (James, 2018): https://www.makeuseof.com/tag/moores-law-silicon-chips/
– (Rotman, 2020): https://www.technologyreview.com/2020/02/24/905789/were-not-prepared-for-the-end-of-moores-law/
Kind regards,
Youri Slok
Nice topic to address there! It is an interesting question to pounder about when thinking about the future of computing. Moore’s law will certainly come to an end, because it depends on the virtuous circle of the manufacturing industry. When a new transistor size is designed and manufactured, the initial investment will be huge. However, if the smaller sized transistor is so small that it has a big-enough improvement compared to the previous ones, people will start buying them massively. In this way, manufactures gain back their investment, hire smarter people, and invest in smaller transistors again. The cycle starts over.
Nowadays, the transistors are becoming so small that they are going to behave quantum mechanically, which shortly means that they cannot behave as predictively anymore. So, going smaller is not possible anymore at a certain size. But, performance can be improved by creating more efficient transistors. This can still achieve double rate improvement. However, we know that efficiency curves flatten out over time. Hence, it becomes harder and harder to make it twice as powerful. Therefore, it will also become more and more expensive. At a certain point, the performance increase does not justify the costs. So, Moore’s law will end.
Another way to look at Moore’s law is not by counting transistors, but by counting performance. It is generally said that the more transistors there are, the more performance it has. This is probably not entirely true, but we can take it as a new measurement to extend Moore’s law. As this is a more general measurement, we can apply this to the upcoming quantum computers as well. Quantum computers already see an exponential twofold improvement in performance per year. All in all, great topic and a lot of things to discuss and pounder about!