The validity of the ley de moore In the medium term, it has been called into question on many occasions, especially during the last decade. The difficulty of constantly refining the photolithographic processes that condition the number of transistors that it is possible to implement in an integrated circuit has caused that this prediction has been about to stop being fulfilled in several occasions. And yet it has not.
The challenges faced by highly integrated semiconductor manufacturers, such as TSMC, Intel, Samsung or GlobalFoundries, to continue miniaturizing transistors and increasing in a practically continuous way the amount incorporated in its integrated circuits are enormous. However, these companies are not the only ones that play a crucial role in this industry.
ASML designs and manufactures the photolithographic equipment used by most of the highly integrated semiconductor manufacturers
They all depend to some extent on ASML, a Philips-owned company in the Netherlands that designs and manufactures photolithographic equipment used by most high integration semiconductor manufacturers. Increasing the number of transistors in an integrated circuit is very important because it allows the necessary improvements to be made in its microarchitecture. to make it more powerful. And also more efficient.
For many decades an essential resource used by microprocessor manufacturers to increase the performance of their chips was to relentlessly increase your clock frequency. This parameter has always gone hand in hand with the improvement of microarchitecture, but, given the difficulty in continuing to increase the frequency, it has ended up giving a large part of its prominence to microarchitecture. And, as we have just seen, for the latter to develop it is essential to continue refining the integration technology.
Here’s what Moore’s law predicts
Gordon Moore has publicly denied on several occasions that the observation he made in 1965 has been attributed the status of law. Actually it’s a prediction fruit of empirical observation, and not a law protected by a meticulous scientific study. Even so, for many decades its observation has become the golden rule used by the microelectronics industry to predict the pace at which it will develop in the medium and short term.
What Moore realized just over five and a half decades ago was that the number of transistors in integrated circuits would double every year, and, at the same time, its relative cost would be drastically reduced. Ten years later, he amended his observation by increasing the time required for this integration technology development to take place, putting it at 24 months, not a year. And since then his forecast has been fulfilled with a more than reasonable precision, as we can see in the following graph.
Innovation has pulled Moore’s law, and it looks like it will continue to do so
Innovation is of paramount importance in the technology industry in general, and in semiconductor production in particular. Without it, Moore’s law would have ceased to be fulfilled decades ago. If we only stick to the last twenty years, we will see that without the innovations that semiconductor manufacturers have introduced in their photolithographic processes, the microprocessors that we currently have would not be possible. And, of course, other highly integrated chips, such as graphics processors, would not have reached the degree of sophistication that they currently have.
Currently we can find integrated circuits produced with 5 nm photolithography in many of the devices that we can buy in stores
At the beginning of the last decade Intel developed the first transistors with three-dimensional structure (known as Tri-Gate), thanks to which it was possible to significantly increase both performance and efficiency of the microprocessors. Another strategy used by semiconductor manufacturers is to use more advanced materials, which, among other advantages, can help combat the electromigration.
And, of course, the photolithographic processes themselves have also developed a lot over the past decades, making it possible to manufacture large-scale chips that benefit from astonishing integration capabilities. In fact, today we can find integrated circuits produced with 5 nm photolithography in many of the devices that we can buy in stores.
In recent years we have witnessed the arrival of many other innovations that have contributed decisively to put us in the position we are in now, such as the use of stressed silicon or extreme ultraviolet lithography (EUV). For this reason, we can look forward with reasonable optimism. In fact, TSMC announced At the end of last April, in 2022, the large-scale production of semiconductors with 3 nm integration technology will begin, and both this company and IBM have revealed that they are prepared to start 2nm chip manufacturing short term. There is no doubt that it is not there but that it is not bad at all.
Images | Intel