November 30, 2015
Moore's Law CPU Power Doubling Rate Slowing

Making computer circuits smaller is essential for making them faster and more powerful. But the size of conducting lines in integrated circuits has gotten so small (14 nanometer in Intel's most advanced wafer fabs) that it is getting much harder to shrink their sizes smaller. Therefore Intel says the rate at which computer power is doubling has slowed to a 2.5 year period.

This matters a great deal for the rate of economic growth. Each doubling in computer power enables more uses of computers to boost productivity in more ways. This slow down in the rate of doubling will slow the rate of productivity increases. Eventually (likely in the next 10 years) the Moore's Law doubling rate will stop. A major driving force of rising productivity which ran for decades will come to a halt. I think

Can advances in computer hardware design still make a big difference once Moore's Law runs out of steam? Possibly quantum computing will step in to boost computing power to far higher levels. I find that hard to judge. What seems more certain: complex algorithms implemented in circuits that take the place of Von Neumann architecture CPUs for an increasing list of specialized purposes.

Another technology seems more certain to become a big source of productivity increases: CRISPR for genetic editing. The fast development of customized plants and animals will likely cause revolutions in agriculture, textiles, drug development, and cell therapy development. A biotech revolution could replace the computer revolution as the next big driver of technological advance.

Share |      Randall Parker, 2015 November 30 07:58 PM 


Comments
bob sykes said at December 1, 2015 5:25 AM:

Moore's Law refers to the density of transitors on a chip. I suspect it actually ended some time ago. With respect to the cpu speed of desktops computers that has stayed around 2.5 to 3.0 GHz for several years now because of heat dissipation problems.

The clock rate of supercomputers continues to edge upwards, but their cooling systems are really elaborate. China's Tiane-2A supercomputer is the world's fastest at 100 Pflops/sec.

http://futurism.com/china-upgrades-most-powerful-supercomputer-in-the-world-hosts-it-in-defense-research-center/

Brett Bellmore said at December 1, 2015 4:25 PM:

For a while, increases in coding efficiency can make up the difference. A good deal of the increased power of computers, at least at the consumer level, (Supercomputers are a different matter.) has been sucked up by coding convenience.

Engineer-Poet said at December 2, 2015 6:19 AM:

The premium will once again be on sharp programmers rather than low-wage equivalents of Lego builders?  That will be nice.

Brett Bellmore said at December 2, 2015 10:04 AM:

If by sharp programers, you mean the ones writing optimizing compilers, sure.

Of course, there's the alternate version of computing, where most of the processing power is a pool of FPLA's, configured on the fly to run the assigned algorithm in hardware. I'd like to see that penetrate the consumer market more.

Somerschool said at December 2, 2015 10:44 AM:

Quantum computing would demand a different kind of measurement than transistors on a chip, but the possibility of adding an infinite number of invisible yet parallel processors would make Moore very happy.

Hal Duston said at December 2, 2015 12:22 PM:

Brett Bellmore said at December 2, 2015 10:04 AM:

If by sharp programers, you mean the ones writing optimizing compilers, sure.

Finding better algorithms and/or pruning unnecessary code is still a necessary skill for a sharp programmer.

Code that is not written or executed will always out-perform code from the best optimizing compiler.

Engineer-Poet said at December 2, 2015 12:29 PM:

Bad programmers can bury the best optimizing compiler in useless cruft.  The sheer size of Windows apps compared to Linux equivalents proves that.

Russell said at December 3, 2015 2:02 AM:

I have always thought Moore's law was an s curve like any other tech. Aviation, for instance. So it naturally would level off over time.

Nicomaine said at December 4, 2015 7:23 AM:

I don't know a lot about transistors or circuits, but didn't Intel just announce something that they found a new way to make transistors smaller?

Post a comment
Comments:
Name (not anon or anonymous):
Email Address:
URL:
Remember info?

                       
Go Read More Posts On FuturePundit
Site Traffic Info
The contents of this site are copyright