Is Moore's law just a special case of an experience curve? [closed]

Question

I recently stumbled upon the notion of an experience curve: http://en.wikipedia.org/wiki/Experience_curve_effects#The_experience_curve . It basically says, that plotting the cost per item over the number of produced items will follow a power law with negative exponent.

http://en.wikipedia.org/wiki/Moore's_law on the other side states, that the number of transistors on a CPU doubles every 18 months.

However if one would plot the number of transistors on a CPU not per year, but per number of produced CPUs (which I can't do, because I do not have the data for this) I guess this would give a power law, too.

In that sense can one say that Moores law is just a special case of an experience curve as also found in other industries?

Answer 1

However if one would plot the number of transistors on a CPU not per year, but per number of produced CPUs (which I can't do, because I do not have the data for this) I guess this would give a power law, too.

No, it would not.

Doubling every 18-months is equivalent to 58.7% YtY growth. Generic version of Moore's law talks about doubling every 2 years, which is still equivalent to 41.4% YtY growth. On the other hand PC CPU sales have flat-lined in recent years. YtY growth for PC CPUs was expected to be 3.2% (units) and 1.6% (value) in years 2013 to 2016. I've haven't yet seen more recent revised forecast, but I'd expect them to be actually even worse.

Answer 2

The answer would be No. - Assuming that the number of CPU's produced follows also follows a power law.

The reason; If n^y CPU's are produced by year y. And if the number of transistors on CPU number n where given by t^n. Then the number of transistors on the CPU produced at the end of year y would be t^(n^y). - So we would expect a super exponetial growth.

While the points indicated by vartec might cast doubts on the question of wether CPU production is still growing exponentially, we should look at the time when the law was formulated.