John Shalf started the session by saying that Moore's Law has been driven largely by improvements in silicon lithography. For the past 50 years, we have seen a doubling of transistor density every 18 to 24 months consistently since the seventies of the past century. With that you get improvements and cost-efficiency and improvements in energy-efficiency that match the improvements in lithography and you also reach improvements in performance, John Shalf explained.
People over the past 50 years declared the end of Moore's Law over and over again to the point that Alan Kay stated: "I declare that Moore's Law will never end and therefore I want to be wrong once". Why does everybody now suddenly think that the end is near for Moore's Law, John Shalf asked himself. As we get down to the 5 nanometer technology and to atomic scale, lithography as a driver for Moore's Law might be at an end, according to some people.
These people claim that all the things that we have depended on in the past for improving performance, started to fail around 2004. This was another one of the drivers behind Moore's Law which in fact is an economic theory. They could not squeeze any more power out, they could not break up the clock frequencies, they could not improve single thread performance substantially. However, industry found a way to continue by doubling the number of cores each generation.
People said: "At least, we still have Moore's Law." This worked with lithography improvements. It was really a hard burden on the software and hardware architecture but the Exascale programme came out of this, John Shalf pointed out. By 2025, when we approximately will get to the 5 nanometer technology node, if we don't see any energy-efficiency improvement with each generation of lithography improvement, there is not really any impetus to continue to go down to 3 nanometers nor 1 nanometer. There will be no more gas in the tank when it comes to improve the performance, John Shalf warned the audience.
However, Moore's Law is an economic theory and although it has been underpinned by lithography improvements, there are other ways to improve things, John Shalf stated. He introduced the distinguished panel of speakers who were invited to demystify their respective technology areas to see what the future will bring.
The workshop is covered in full in five articles: