Moore's Law
Dean
Moore's law states that computers double in power and go down in price about every 18 months. Although that sounds ludicrous, that law has held true for over 40 years now and shows no sign of slowing down today. As far back as 1980, a 1 megahertz computer cost around two thousand dollars. Today a computer running at 3 gigahertz--3,000 times faster--sells for well under a thousand dollars. And that's just the clock speed, that's not counting the fact that those 1 mhz processors only processed 8 bits of data at a time, whereas modern applications process 32 or more bits at a time. The machines now are roughly 15-20,000 times more powerful. In only 25 years.
If Moore's Law continues to hold, well, do the math. Today you can buy a 3 gigahertz computer for under a thousand dollars (EXAMPLE).
What will 20 years bring? 20 years divided by 18 months is 13.3.
3 gigahertz 2^13d power is 30,257 gigahertz*. There'll be storage and memory amounts that have multiplied that far as well. All for well under a thousand dollars.
What will we be able to do with that kind of computing power? The mind boggles. But don't be foolish and say "no one will ever need that much." The more you can do, the more you want to do. My first hard drive had ten megabytes of storage and cost about $600. Today you can buy USB memory drives with 128 megabytes of storage for under $20. Pictures, songs, movies, a running video recording of your entire life, and more. There's no end to what people will want to do, or be able to do.
It's fun to think of the top side of that, but the dangers will be there too, and we will see it on our watch. Imagine what you can do with gene splicing equipment in your own home.
* - Corrected
* Update * An objection is made that processors can't keep going up in clock speed ("hertz") forever. But this has long been true; indeed, everybody who knows how they work acknowledge that the "hertz" rating is really just a very rough shorthand for speed. Parallel processing, wider buses, and new materials mean that processors are evolving rapidly in their power. Chances are that by 2025 we won't be referring to the "hertz" rating at all anymore, but the equivalent power will have doubled many times by then.
Related Posts (on one page):
"Imagine what you can do with gene splicing equipment in your own home."
Three years ago I bought a top-of-the-line workstation whch contained two Xeons. They run at 2.4 ghz. Today the fastest Intel chips are only about 50% faster than that.
There will be further speed increases in future, but not at the rate we saw historically. They've run up against fundamental physical limits: the speed of light, Planck's constant, and the size of atoms.
http://arstechnica.com/articles/paedia/cpu/moore.ars
Basically, it's not really a solidly stated "law". The closest it goes to is transistor densities, but that doesn't mean "computing power".
Moore's specific prediction had to do with the density of transisters. We could see the writing on the wall on that some time ago. But the changes in material, and the increase in parallel processing, and the decrease in power needs, and so on, have corresponded to ongoing increases in raw computing power anyway.
Hardware engingeers seem to have taken Moore's idea of doubling power about every 18 months to heart, and the software engineers have taken it to heart as well. I vividly remember that in the 1990s we were being told that Moore's Law had stalled out, or would soon: motherboards could not keep up with the processors, voltage requirements were becoming too high, and the limits of silicon were being reached. I was told even then that Moore's Law had finally stalled. Since then we've gone from 50-60 Mhz processors stuck on 10-20 Mhz busses to busses in the hundreds of megahertzes and processors in the giga hertzes.
I also recall as far back as the early 1990s being told that parallel computing, what amounted to multiple computers in one CPU, was mere pie in the sky fantasy that looked promising but no one knew how to use such hardware, to it being simply built into the chips; it's now an everyday thing and barely anyone even groks that it's happening.
The truth is that the "hertz" rating stopped having a precise meaning quite some time ago. The added complications render that manifest: motherboard speeds, parallel processing, anticipated processing, multiple instructions processed per cycle, and so on all add to ongoing increases in power without necessarily boosting the "hertz" rating.
Oops. I was obviously tired (I wrote that just before bed.)