The most widespread misconception around Moore’s law
Drawing from an original article of 1965.
What will happen if you ask almost anyone, from readers of this blog to Wikipedia, what was Gordon Moore’s observation about the development of electronics? The answer will be a doubling every few years of everything and everything, starting from the number of transistors and ending with performance.
But all this is just a consequence, and how it really is - let's understand.
* I want to say right away that I will not write about what has already been said more than once (a dozen?) Times. I will try to be brief and to the point.
Original article of 1965It was written at a time when no Intel company existed yet, and its future founders were busy looking for ideas for their business. In particular, the driving force that can make electronics truly massive. And the main obstacle for this was, first of all, the price.
Accordingly, the initial observation was how the cost of one transistor behaves depending on their quantity and production technology.
For small (by the standards of a particular technology) devices, the cost of one transistor is almost inversely proportional to their number. But, with an increase in the number of transistors, the crystal area grows, the yield fraction of suitable crystals decreases during production, and, accordingly, the price rises.
It turns out that for each production technology there is a certain level of complexity of the devices (expressed by the number of transistors in their composition), for which the use of this technology is most beneficial.
The transition to a new process technology is achievable when the cost of the transistor according to the new design norm is significantly lower than for the current technology. Here a forecast was made about a tenfold reduction in the cost of the transistor when changing the process technology.
As for the number of transistors in the device, for which production is the most economically profitable, according to forecasts, when switching to a new technology, it really doubles.
But, firstly, this is a consequence that can be obtained by constructing the minimum cost points in the coordinates “year” - “number of transistors”.
Secondly, this forecast has not been fulfilled in its original form for a long time - it should be adjusted taking into account the area of the crystal itself. Several recent technological processes, instead of doubling the number of transistors, bring only a marginal increase in their number and a significant reduction in the area of the crystal itself.
In fact, now that a new microprocessor is being developed, the main limiter is not what capabilities it should have or what productivity should be. The main limiter is how much it should cost, from where everything else follows.
Moore's Law has been and remains the law of value. And many things would be clearer if people remembered this.
Did you know about this side of Moore’s law before?