yes, moore law talk of size of transistor, but that is linked to speed, complexity, RAM size...
there are many different moore laws and when they interact with dimensions in real world it created big problems. for exemple disk and RAM memory capacity grow faster than the I/O to talk to the disk. this required cache with more and more efficiency, and processor which speculate more and more (supervectorial)... or who pipeline hundreds of cycles (MTA architecture), with memory hierarchy higher and higher same for the processors that grow faster than it's IO, and than heat dissipation capacity... One of the problems is as said here the radius of synchronicity... this is why processor get more and more asynchronous, parallel, pipelined, and that programmers have to change their algorithms to use parallelism, despite more instructions used (Montecarlo, diagonalizations, gradient solvers...) there is also a moore law for the size of silicon wafer a limit of expomential laws is the financial requirements which follows a similar moore law. anyway as Mats explain there is breakthrough that allows to get out of those limits by changing the rules. for example in companies there is an interest in concentrating resources, but it creates problems. new management allows loosely coupled actors to work together benefiting from the size effects, while not being hindered by big corporate culture... see for example Uber or Blablacar or AirBnb, or simply internet compared to NBC news. you can see that in real problems. moor is only a detail for most progress. In a course someone talk about the speed of eliptic curves solving. Moore law is a minor effect in the huge increese of speed, which was pushed by huge improvement is mathematic. Same for cryptographic problems like hash, cypher, primes... 2015-01-27 13:48 GMT+01:00 John Berry <[email protected]>: > BTW AFAIK Moores Law isn't that speeds increase but that transistor > densities increase. > > On Wed, Jan 28, 2015 at 1:33 AM, Bob Higgins <[email protected]> > wrote: > >> What James says is true about the radius of connection. However, two >> things have been driving that radius smaller - smaller gate size and chip >> stacking. We all recognize that making the transistors and the gates >> smaller decreases this radius, but what is not widely recognized is chip >> stacking technology is becoming more common. The issue with chip stacking >> is the heat dissipation. This was addressed by IBM, for example, using >> liquid cooled systems and stacking years ago. However, once the IC power >> is reduced, chip stacking becomes very practical. It is currently used to >> stack memories on top of processors in a lot of consumer devices. >> >> Going up provides a lot of opportunity for increased performance by >> adding complexity without substantially increasing the radius of >> connection. There is presently a lot of headroom in this technology for >> additional Moore's law advance. >> > >
