Every year, as computer hardware becomes ever more capable, we are able to make use of increasingly sophisticated software.
This increase in power, meaning that our coders can produce more effective products, has followed an approximate curve, known as Moore’s law. But without a radical change in hardware design, this must soon come to an end.
Moore’s law is named after Gordon Moore, the American technologist who co-founded Intel in 1968. While still working at Fairchild Semiconductor in 1965, Moore noted that advances in hardware meant that the number of components on a computer chip were doubling each year. This was changed to doubling every two years in 1975, an observation that has broadly held true ever since. This phenomenal growth in capability has enabled the transformation of software from relatively simple applications to the powerful products available today.
The change is often shown on a logarithmic plot, to demonstrate that the increase in power has continued on a doubling path, but this makes the exponential growth less obvious as becomes clear when you see the real values plotted:
Figure 1- From My World in Data, data source Karl Rupp
However, the word ‘law’ in Moore’s law is misleading. In science, a law is usually an unchanging aspect of nature. But all Moore’s law can ever be is an observation of what has happened so far. And we know that there is a limit to how much more capacity can be crammed into a single chip of any manageable proportions. The main way that capacity has increased is by shrinking the size of individual components like transistors. But the electrons that move around the inside of a chip are quantum particles.
Unlike familiar objects, there is a probability that quantum particles will ‘tunnel’ through insulated barriers that should stop them in their tracks. This can be beneficial – it’s how, for example, a memory stick or a solid-state drive (SSD) works. Normal computer memory loses its contents when the power is switched off. But in the flash memory used in these devices, the memory store is insulated to keep the electrons in, with the contents modified by using this quantum process of tunnelling that enables the particles to jump through the barrier.
Unfortunately, when it comes to processors, as the insulated gaps between components continue to get smaller, the barriers are now reaching the size where such quantum tunnelling is inevitable, meaning that the steady march of Moore’s law could soon be challenged.
If hardware manufacturers are to continue to enhance their products, enabling coders to continue developing of increasingly powerful software, whether it be for the next generation of office tools or massively powerful artificial intelligence, it seems likely that we will have to change the physical basis for our chips from the longstanding standard of CMOS (Complementary Metal-Oxide Semiconductor) technology to new approaches.
One major opportunity is likely to be through the deployment of 2D materials. These are ultra-thin substances like graphene and boron nitride, which are essentially crystals that are one atom thick. Graphene remains the best-known of these substances, isolated for the first time at the University of Manchester in 2004 by removing layers from the form of carbon known as graphite (used in pencil leads). This initially involved the unlikely method of using repeated applications of sticky tape.
Because of their single-atom thickness, these substances have unique electronic properties – it seems likely that by combining layers of different 2D materials it should be possible to produce significantly smaller electronic components, enabling hardware manufacturers to effectively keep Moore’s law alive after the existing technology can no longer continue to be shrunk.
The law that has enabled our continued development of more capable devices and software may well be able to survive for years to come thanks to these remarkable new substances.