Moore's Law?

[jasong]

I'm wondering everyone's opinion on the state of Moore's law. Obviously, the wording has gone through so many incarnations that I'm not even sure of the exact quote. Be as technical as you want, but I'm basically asking how long people think this seemingly breakneck pace of increases in technological complexity shall continue for cpus, RAM, hard drives(yes, I know Moore's law is stated differently for some tech), flash drives, and anything else that is advancing extremely quickly.

[Xyzzy]

Well, first of all, it isn't a law. It is an observation.

[jasong]

Originally Posted by Xyzzy

Well, first of all, it isn't a law. It is an observation.

Well, yeah, that's true. I wasn't trying to suggest it as an absolute. As a matter of fact, the way I asked the question, your answer was kind of wise-assy. ;) (Welcome to the club)

[rgiltrap]

Originally Posted by jasong

I'm wondering everyone's opinion on the state of Moore's law.

Moore's Law certainly has a few iterations left, 45nm, 32nm, 22nm are all in the pipeline but there will obviously be limits set in at various stages. One of those practical limits has already been hit... POWER consumption.

Up until recently the silicon real estate was thrown at making longer pipelines and cranking up the clock speed, which in turn required larger caches to feed that pipeline. Transistors were effectively free to use, as more became available you just threw them at the processor in this way (NB: very much an over simplification). However, the power envelope problem has forced a change in design to multi core cpus.

The next practical limit is whole of system performance, cache only goes so far. We have already seen the addition of the memory controller onto the Athlon processors and with AMD's purchase of ATI they have already indicated that alot more functionality will move on the cpu (graphics, networking etc).

Moore's Law may hold true for a long period of time or may start to slow either way the problem is now what is the best use of the available silicon, pipelines, fpus, cache, cores, other functionality?

CPU design is changing to System on a Chip (SoC) design very very quickly as such not every piece of silicon will be focussed on faster prime hunting

[cheesehead]

Originally Posted by Xyzzy

Well, first of all, it isn't a law. It is an observation.

... from which many folks have formed extrapolations, on which the same or other folks have based predictions ...

[Xyzzy]

... from which many folks have formed extrapolations, on which the same or other folks have based predictions ...

Still ain't a law.

[cheesehead]

I wasn't disagreeing.

[wblipp]

Originally Posted by rgiltrap

POWER consumption. ...
more functionality will move on the cpu

We've been here before. I remember of the move from 5V to 3.3V chips and the integration of the stand alone math coprocessor, both in the 1980's IIRC.

[jasong]

If we do hit a wall, when and what do you think the wall will be? And has anybody considered the possibility that when we're forced to abandon silicon, we might suddenly find ourselves getting a shot in the arm in terms of performance possibilities because of new, never before used, technology?

[davieddy]

Originally Posted by Xyzzy

Well, first of all, it isn't a law. It is an observation.

Yes.
Like Ohm's Law, or Hooke's Law.

[Xyzzy]

We haven't had time to read this, but here is a link that popped up today:

http://arstechnica.com/articles/paedia/cpu/moore.ars