Slashdot Mirror
Gordon Moore: Moore's Law is Dead
Golygydd Max writes "Moore's Law will not hold forever, claims Gordon Moore.
In a Techworld article, he points out the limitations of the law, in particular, the limitations as we approach the size of atoms.
He helpfully explains, however, that the law will hold for a few years yet." Still, sticking around for forty years is pretty impressive.
I wish I could mod the Wikipedia article up.
"This textbook contains material on Moore's Law. Moore's Law is a theory, not a fact, regarding the scaling of computer processing power. This material should be approached with an open mind, studied carefully, and critically considered."
I agree, 40 years is actually pretty short. Most common math was proven hundreds to thousands of years ago. A good portion of physics was known a few hundred years ago. A good portion of chemistry has been around for about 150 years.
What is impressive: he predicted the growth would follow the trend it did, in an area that hadn't really been well-established.
Which leads to a second dilemna: since Moore was heavily involved in the industry that the law describes growth in, did Moore's law follow the natural growth, or the growth match Moore's law because industry decided to follow the law?
Though of course your post is a joke, the answer is no. Moores law itself wasn't just a number that he pulled out of his arse, but a serious study of transistors and statistics. But back then approaching the size of the atom with a transistor must have seemed a *very* remote idea. As the summary says holding for forty years is an achievement in itself.
That said CPU power isn't just a measure of transistor density anymore (it was at least in Intel propoganda for a while), as you can see with the dual core and 64 bit developments. There's still plenty of juice left to be squeazed out of the current design before it's squeazed out.
No it can't, because we still don't understand how the brain(s) work, because the neurons ain't the only thing working in there, ...
The best thing we can do is throw random "computing equivalent" numbers and check if we're there right now
And these random numbers are modified every other morning...
"The way we can tell it's C# instead of Haskell is because it's nine lines instead of two." -- wadler
I know the poet's version of the law, that the number of transistors doubles every year, but why do people make such a fuss about it other than the fact that it's a nice little prediction? That is: Ok, we've observed this dynamic; does it have any practical implications whatsoever?
As other posters have noted, Moore's law is about transistors. Kurzweil in his book uses a much more liberal extension of the law which allows him to look at technological development from the stone age through to speculations about the far future. Obviously they didn't have transistors in the stone age. They didn't even have tubes.
Loose lips lose spit.
When Moore first proposed Moore's law, it had nothing to do with processing power. He was making a pretty ambitious prediction about transistor density on ICs. Then Moore's law was about memory density, then later about processor speed, then finally about "computing power".
Moore's original law was more insightful at the time, if more narrow, than the current one.
Socialism: a lie told by totalitarians and believed by fools.
Nothing against you, but how the hell did this get rated insightful?
Moores law is about transistor density, not speed.
Plenty of problems here.
1 - An N-state transistor takes roughly N units of space and N units of power in exchange for log(N) bits of data. The natural number (e) is the theoritical ideal number of states for a transistor, and anything above that is less than ideal.
2 - Computational power is limited by surface area, not volume. The thicker transistors are packed, the more heat is made, and the slower they have to run.
3 - Exponentials grow really, really fast. Moore's law in particular also has a very high constant, doubling every 1.5 years. I doubt Plank's constant will halve every 1.5 years or the Earth will keep doubling every 1.5 years to make room for our ever shrinking transistors or ever growing chips just to be friendly. Even if we can manufacture ever more powerful chips at that rate, the heat output will eventually overwhelm our power plants and fry our planet.
4 - Quantum tunneling only hurts for transporting current. It means that the our chips are descending into randomness and chaos as they shrink. I fail to see how this can be exploited for moving electricity down wires, when you want the electrons to stay in the wire, not tunnel to adjacent wires.
Add Hyperthreading to that list. It's actually pretty brilliant. Basically two chips that share execution resources on a single die. Twenty years ago, you couldn't put two virtual processors on basically the same die size as one standalone processor. As feature size gets smaller, you can add lots of extra goodies to make sure that more of your transistors are doing something useful more of the time.
Today's Sesame Street was brought to you by the number e.