In March 2023, Gordon Moore passed away at a very old age. He, an engineer and co-founder of the Intel group, came up with “Moore’s law”. According to this, the performance of microchips, or to put it more correctly: the complexity of integrated circuits that can be produced at low cost, is doubling at regular intervals.
magazine article
Published he had his thoughts on April 19, 1965, 58 years ago today, in the magazine “Electronics”.
He was right. And yet since then there has been one thing in particular: uncertainty: about how long the doubling period is and about the expiry date of the law. The chip industry needs knowledge of both – and actually also all the associated industries from the computer hardware group to the manufacturer of modern washing machines to the video game developer in order to be able to plan to some extent.
In 1965, Moore assumed a doubling time of just one year. A speech by him in 1975 is documented, in which he then speaks of two years. In fact, the values always varied somewhat, depending on certain innovations, for example. And then you first have to agree on what you actually want to measure and count, and whether you consider the number of circuits per unit area or per chip to be the decisive factor when you use the abstract concept of a “doubling of complexity”. want to determine.
Cleverness and finiteness
In addition, some factors not only vary, but only change in one direction over time. In the 1970s, for example, engineers were already reaching their limits with what Moore called “cleverness” and by which he meant the ability to arrange components and transistors in a space-saving manner.
All of this ultimately led to the fact that there are now a few different versions of “Moore”, depending on your point of view. In any case, the value for the most popular of these variants is currently around 20 months, and 18 months are often given.
Moore himself predicted an end to the validity period as early as 2007. He gave his law another 10 to 15 years. As so often in similar cases, from the slow drying up of oil wells (“peak oil”) to agricultural yield capacity, this also seems to be dragging on a little longer. Prediction currently: sometime in the 20’s.
That it will actually happen, although a lot is still possible when it comes to miniaturization, is considered certain. One explanation for this is a purely physical one: at some point, adequate cooling of the increasingly tightly packed, current-carrying and heating-up transistors will no longer be technically possible.
Read all the episodes of the “Tagesrückspiegel” column that have been published so far here.
