"Eternity is long, especially toward the end: another short story of Moore's Law"

I would have gladly attributed the first part of the title of this editorial to Michel Audiard before learning that the author of this quote is none other than the Austro-Hungarian writer Franz Kafka. You will therefore set your reading level at the height you wish, somewhere between the lightness of the former and the often sinister universe of the latter.
Alternatively, a variation of this title could have been "No exponential is forever: but forever can be delayed" whose authorship goes to Gordon Moore (Intel), eponymous author of the law discussed here. But who enacted this law that everyone thinks they know but very few manage to state its content? Moore's law is the story of the formidable industrial revolution that originated in the miniaturisation of integrated circuits, which today contain billions of transistors. It is also the law that governs the digitalisation of our society. Here is a short history...

The origin of Moore's Law...

It was on the basis of a small set of data, while he was an employee of Fairchild, that G. Moore had empirically formulated around 1965, that increasing the integration density by a factor of two at regular intervals would allow minimizing the manufacturing cost per unit transistor with a gain in frequency performance. Starting with a 32-component integrated circuit in 1964, Moore came back to support his projection in 1975 at the IEDM (a conference driven by a strong industrial audience): a 65,000-transistor memory circuit from Intel's manufacturing lines confirmed his prediction of exponential growth. It was in this context that Moore's metronome started to beat with a roadmap that made the increase of integration density and performance on the one hand, and the reduction of cost/bit on the other hand, resonate.
At the same time, R. Dennard (IBM) theorized the scaling law of the MOS transistor, which demonstrated that its power density remained constant with the reduction of its dimensions (and at the same operating frequency!). Dennard's work confirmed the promise of a doubling of the integration density and a 40% gain in frequency for a homothetic reduction of √2% in the dimensions of the transistor and an increase in the channel doping in the same proportion.

Moore's Law became a self-fulfilling prophecy, and the semiconductor industry devoted itself to anticipating the development of new processes and equipment, especially lithography. Never before has an industry had such a well-defined roadmap, with a trajectory that sets both the technical and economic rules of the game.

The road was full of challenges and technological obstacles: we knew where we were going without having all the keys to get there, but we were going! Aware of the challenges posed by the continuation of Moore's Law, the world semiconductor industry agreed in 1993 to formalize the ITRS roadmap (International Technology Roadmap for Semiconductors), which allowed precise quantified objectives to be set on a quasi-irrevocable schedule. Thus the first ITRS technology node was positioned on a pattern size (the gate length in first intention) of 0.5 μm, projected 300,000 gates per chip, anticipated DRAM and SRAM memories of 16 and 4 Megabits, respectively, all for a cost of $4 per cm2! The regular interval, mentioned above, was now set at three years to progress from one technological node to the next...

To subscribe to the newsletter, click here

To unsubscribe to the newsletter, click here