A Century of Moore’s Law

semianalysis.com

5 pointsneatze4 comments

Transistor scaling being a notional scale now. So in some sense, when the nm stopped being literally density-informing, Moores law became old Moores almanac.

As Ray Kurzweil points out, Moore's Law isn't as interesting to most of us as the overall exponential progress of computational power that combines advances in both hardware and software. The hardware-only aspect should only be of interest to semiconductor manufacturers (or lazy software developers).

It's informing of power budget: battery life and heat burden, both of which consumers care about.

Frequency, maybe less so. But the clock speed asymptote had an effect on core count and core count has driven most of the effective visible improvement in performance against expectations of camera CCD density, image quality, and consequent os and app complexity.

The exponent is changing.

But he's not wrong.

Better algorithms improve battery life and heat burden too so I don't see how focusing on hardware is better for the consumer.

Reply on news.ycombinator.com