Moore’s Law was not dead, was on a spree

Gordon Moore spent 20 years ashamed of the law that bears his name. In 1965, the co-founder of Intel made an estimate based on empirical observations on the semiconductor market. According to his data, the number of transistors per unit area would double every year.

In 1975 revise the remark to indicate that the term would actually be 24 months, and suddenly that Moore’s law became not a forecast, but rather in a script (or challenge) had to evolve how industry semiconductors. However, there are many factors that we are preparing to stop abides by the law: the future of semiconductors will soon have other objectives.

A law that has given the technological evolution for decades

The industry assumed that rate in the scale of integration was an acceptable target and achievable. Two bodies, one basically American, the Semiconductor Industry Association, and other global, the International Technology Roadmap for Semiconductors – took it as perfect metronome developments in semiconductors, and for decades have witnessed the frantic race in which every year we saw how lithographic processes improved to allow more and more reductions in manufacturing scales.

To these reductions were accompanied improvements in material or in the format of the transistors (came the three-dimensional), allowing one year after another quiet breathing: Moore’s Law was still in force and we could ensure that our processors would be more and more powerful, many have more million transistors in each new iteration.

Things have become more and more difficult in recent years: The photolithographic processes have still room for improvement technology – the extreme UV wavelengths promises 13.5 nm as currently wavelengths of 193 nm-used but still there will be a time when the transistors have to be so small (10 atoms wide if we were to stop manufacturing 2nm instead of the 14 some current processes) that guarantees a solid and efficient operation would begin to be difficult to give.

Times change

This has been linked to current events and trends such as the rise of connected devices, the Internet of Things that promises to make our lives a utopia (or dystopia) of hyperconnection has changed the paradigm: no longer need a scale integration amazing to have more powerful processors need more integration of different types of components. The SoC today concentrated CPU, GPU and some components (smartphones are well aware of this) will be even more ambitious in this area in the future.

The Internet of Things demands precisely this integration of sensors and components in one package, and that seems to be the challenge of the future. The ITRS publish its next roadmap in March, and its leaders have already advanced in the journal Nature that the objectives change: no longer subject to the law of Moore that was beginning to choke manufacturers.

Even the powerful Intel had to change their plans and delay the launch of Cannonlake 10 nm, which will not arrive until 2017. Here Intel has been the victim of Law Rock, reverse of Moore. Under the law, the cost of a chip fabrication plant doubles every four years, and thus the costs involved in research in new materials and added integration technologies before implanting them in these factories.

You may also like to read another article on Lab-Soft: Intel fixes bugs in the processors Skylake

Moore can relax a bit

Although the intention of the industry is to loosen up a little knot of his tie and migrate to objectives not strangled, developments in manufacturing technologies continue to exist and continue to occur, and in fact there are many open doors in the future that could rethink again the validity of Moore’s law or at least a review of the terms in which they must comply.

In fact as we say Intel will continue preparing the jump to 10nm over the next two years, but it is unclear when it will be able, or even whether it will be able-to take the next leap in lithographic processes according to their roadmaps would place 7 nm in the range of 2018 and 2019 leaving behind silicon and that of 5 nm in 2020 and 2021, which are already preparing the processes named as nanowire FET, an evolution of the existing FinFET.

Then experts say Moore’s Law no longer be met, but even then only see a little more relaxation. The use of new materials – spent years talking graphene – superconductivity or also promised revolution of quantum computing could re-validate this theory and even review it downwards. Maybe Moore’s law is currently before a precipice, but we must also keep in mind that this law still saved a few aces up its sleeve. In here you can find more tech news and reviews

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.