• Question: How many years will moores law remain true for?

    Asked by on 15 Jun 2020. This question was also asked by .
    • Photo: Andy Smith

      Andy Smith answered on 15 Jun 2020:

      Moore’s law relates to the number of transistors on a computer chip roughly doubling every two years. There’s two ways to continue this trend – you can make the chip itself bigger, or you can make the transistors smaller. The laws of physics mean that we’ll be reaching what we believe to be the limit of making transistors smaller in perhaps 10-15 years. And making the chips bigger isn’t without its problems too – the bigger the chip, the more chance there is for a manufacturing defect which means the chip won’t work.

      This is unlikely to be a real problem for us, however. We can continue to build more powerful computers by plugging in multiple chips, or by connecting multiple computers together into a super-computer. We’ve been doing both these things for many years already.

      Also, in many cases we don’t care too much about raw computing power. Instead, we care about power efficiency, especially for mobile devices. Afterall, there’s no point in you having the most powerful mobile phone if the battery only lasts 30 minutes!

    • Photo: Anar Yusifov

      Anar Yusifov answered on 16 Jun 2020:

      -1 year… Sorry, couldn’t resist. Actually it makes less and less sense nowadays. Architectures are going crazy in terms of how they stack processors. Vendors creating more specialized cores and setups. All in order to sell more, while they still can. Beware of Quantum – in 5 years it might be Neven’s Law time…

    • Photo: Steve Williams

      Steve Williams answered on 16 Jun 2020: last edited 16 Jun 2020 8:47 am

      Moore’s law isn’t a real law despite the fact that it has held mostly true for many decades now. It’s more like a rule of thumb which gives really good approximations. Moore’s law as I understand it relates to what is known as the Von Neumann architecture which is basically the architecture of most modern computers but without some of the modern refinements. There is also another so called law, the law of diminishing returns. This law in as much as it is a law predicts that it requires more and more effort and ingenuity to get further gains of performance out of a given technology. The further you increase the performance the harder it is to increase it the next time. This to a large extent is where our modern computer are now. So for a rough example, you double the number of transistors on today’s computer and you get a 20% increase in performance. In two years time you double the number of transistors again only to get a 15% increase in performance. It’s starting to look like we are approaching the near absolute limit of the Von Neumann architecture. So to move computing power to great heights we need a new architecture or technology. That new approach is likely to be Quantum computing which promises to make huge advancements in performance. Let’s see what happens next.

    • Photo: Oliver Gordon

      Oliver Gordon answered on 16 Jun 2020:

      I like this question! Of course Moore’s law isn’t a “law”, as much as a coincidence in how technology has progressed over a longer period of time.

      I’d say we’re pretty much at the end of it, to be honest. You only need to look at Intel – they’ve been unable to shrink their processor size for several years now. There’s also the question of what comes after Silicon – why would Moore’s law apply to graphene and quantum processing, for example? There’s a group in Canada we work rather closely with who have even been able to run computer code by picking up and placing single atoms – quite cool!

    • Photo: Katy Brown

      Katy Brown answered on 16 Jun 2020: last edited 16 Jun 2020 11:04 am

      We also often use Moore’s law to relate to the cost of genetic sequencing – which reduced in price by half every year from 2001 (when the first human genome was being sequenced) until 2007 (when “next generation sequencing” was introduced).

      Since next generation sequencing was introduced, the reduction in cost has overtaken Moore’s law by a long way. Sequencing a single genome in 2001 cost $100 million (USD), in 2007 it cost $10 million and it now costs less than $1000 – it is 10,000 times cheaper than 13 years ago! Some companies are now close to being able to sequence a genome for less than $100.

      This graph shows the pattern –

    • Photo: anon

      anon answered on 16 Jun 2020:

      Very good question! Actually we’re now starting to talk about “the end of Moore’s Law”, and Gordon Moore himself thinks the law will end by 2025. Making things smaller and smaller is a good method up to a point – others in this thread have mentioned that this means the devices use more power, and heat up more. Another issue is that something known as “quantum tunnelling” comes into play – this is quite a strange thing where electrons (the small particles which move to make electricity, and which make all circuits work) suddenly behave like waves instead of particles. This lets them “jump” (or “tunnel”) from one very tiny wire to another one near it, meaning that a wire which had (and was meant to have) a value of ‘1’ suddenly has a value of ‘0’, and the other way around for the wire which the electron has jumped to. This means nothing would work at all! Of course, scientists and engineers are working on new materials which stop this from happening, but there will be a point where we just need a new technology to come along.

    • Photo: Emma Wilson

      Emma Wilson answered on 18 Jun 2020:

      Moore’s law is an observation and projection of a historical trend, rather than an actual law of physics. It just so happened that there was a pattern in the data until eventually there wasn’t.

      Sorry to go off-topic, but the concept reminds me of a graph that my old statistics tutor showed me back during my degree. It was a regression analysis of the winning times of the fastest male (blue) and female (red) 100m sprinter in the Olympics.
      Without worrying about the statistics behind what a regression analysis is, the graph predicts that in the year 2252 a woman should be able to win the 100m sprint in under 7 seconds. This is highly unlikely, and is a misuse of regression analysis.
      Here’s a link to a blog post explaining it in more detail (the graph is in there too) if you are interested: https://towardsdatascience.com/will-a-woman-be-the-fastest-human-502be89e0060

    • Photo: Jane Kennedy

      Jane Kennedy answered on 22 Jun 2020:

      Already lots of great definitions here of Moore’s Law which I don’t need to repeat. The side effect of this doubling of transistors is that computers and computer chips have also been getting twice as fast every two years. As others have said, we are reaching the limit of how much smaller transistors can get anymore, so since we can’t have more of those, computer scientists have been looking at other components (known as accelerators) that can be used alongside the main processor chip in order to make computers faster and more powerful. A common example is a GPU, which will often be found in gaming PCs and video game consoles to help with the graphics, which the processor would struggle with on it’s own. Other posters have mentioned quantum accelerators, which may be another option in future.

      The thing about adding more components to a computer, is it makes them much more difficult to program. Even a program that already works will need changed to run on more specialised devices, which isn’t always easy or practical to do. On the other hand, if you have a super-powerful computer and only a small problem to solve with it, there’s not much benefit to having all that compute power. So even if it’s possible to make computers faster, it may not be something that’s very useful if we lack both the problems, and the people with the skills to make them work.

      This question is very timely, as the Top500 list of the fastest computers in the world was decided this week! Number one is a new Japanese supercomputer called Fugaku, it has a massive 7.3 million computer cores (my PC here has 4) and was measured doing more than 400 MILLION BILLION sums per second. See the full list here: https://www.top500.org/lists/top500/2020/06/

      The website has lots of interesting stats, there is this diagram that shows Moore’s Law vs actual performance for the fastest computers in the world on the bottom of this page: https://www.top500.org/statistics/perf-devel/

      You can see that even though Moore’s law isn’t a true law, total and average performance for all the top 500 has actually followed what you would excpect, up until about 2015 when performance has started to flatten-off and dip below the Moore’s law projection.

Comments