Is Moore's Law Finnished

clearz

Moores law states that computing power will double every 12-18 months

http://www.intel.com/research/silicon/mooreslaw.htm

I bought a CPU in January 2002. It was a 2Ghz P4 and the fastest then was 2.2 Ghz. It is now August 2004 and the fastest is 3.6 Ghz that is hardly doubling every 12-18 months. According to that law we should be over 5 Ghz by now.

jesus_thats_gre

Simple answer, sort off. You are applying that rule to consumer products though!! Maybe processing power has doubled in the past 18 months in the server market?

radiospan

Clock speeds seem to have hit a sticking point all right around the 3Ghz to 3.4GHz mark.

Although I don't know whether or not Moore's Law applies specifically to clock speed, or to some other benchmark.

Billy Kovachy

The law states the doubling of components on a circuit as you can see in the graph on the web page you supplied a pentium 3 has 24million transistors then in 2000 the pentium 4 has 42million transistors.The manufacturing process has double the components and due to supply and demand the manufacturing cost for a pentium 3 has halfed in a year.As the paper says

"the cost advantage continues to increase as the technology evolves.."

your looking at the wrong aspect and wrong area of cpu sales,for moores law look at the bottom end chips and the number of stuff you can fit on them.Actually a microwave would be a better place to look at, have you not noticed through the years more and more crap with buttons and timers thats the effect of Moores law.

Tom Dunne

clearz wrote:

Moores law states that computing power will double every 12-18 months

http://www.intel.com/research/silicon/mooreslaw.htm

I bought a CPU in January 2002. It was a 2Ghz P4 and the fastest then was 2.2 Ghz. It is now August 2004 and the fastest is 3.6 Ghz that is hardly doubling every 12-18 months. According to that law we should be over 5 Ghz by now.

Not necessarily. Yes, the jump in speeds has slowed, but the real difference is in the number of transistors, as Billy Kovachy has pointed out. In very simple terms, the more transistors, the more functionality. The latest processors have hyper threading, which will provide some parallelism in the execution of instructions, they also have more cache inbuilt on the die itself.

Check sites such as Toms hardware (no relation!) for performance comparisons. You should see a performance difference that could not be explained by raw clock speed increases, but more to architectural changes.

stevenmu

Somebody else worked out that every now and then, I think every 8-10 years roughly, the increase slows off for while before new technologies come on stream. When they do, the increase they bring makes up for the slowness of the previous years, when dual-core 64 bit chips come out in the next 2 years or so that's exactly what's going to happen.

Draíochta

Intel are moving away from naming their processors according to clock speed, they are going to follow AMD and use names and mumbers, like prescot 4000 and Xeon 4000, (made those names up) I always thought that Moores law was famous because it will fail, as the size of the transistors goes down, they get too small and cannot cope with electronic interferance.

"Will we ever have the amount of computing power we need, or want? If, as Moore's Law states, the number of transistors on a microprocessor continues to double every 18 months, the year 2020 or 2030 will find the circuits on a microprocessor measured on an atomic scale. And the logical next step will be to create quantum computers, which will harness the power of atoms and molecules to perform memory and processing tasks. Quantum computers have the potential to perform certain calculations billions of times faster than any silicon-based computer. "

http://computer.howstuffworks.com/quantum-computer.htm
Quantum computers, an interesting concept.

Capt'n Midnight

241-248GHz is an amatuer radio frequency in many countries and on application here. http://www.comreg.ie/_fileupload/publications/ComReg0477.pdf

so there is another factor of 100 in the raw speed of semiconductors. Whether this can translate to silicon cpu's is a different matter.

also since multiple cpu's are starting to be supported by OS's and since the pentium CPU's have more than one ALU inside them even if silicon hit a hard limit in speed , parallel processing has a long way to go. eg: The connections machine had 65,536 or so pentiums in parallel

compare moores law to RAM sizes / Flash disks and even hard drive capacities...

Draíochta

Capt'n Midnight wrote:

so there is another factor of 100 in the raw speed of semiconductors. Whether this can translate to silicon cpu's is a different matter.

I don't think that the speed of a transistor is at question, it is not just the speed at which one transistor can be clock but how quick you can clock millions of them without failing due to transmission line problems, i.e. interference from nearby transistors.
Hyperthreading and dual/multi processors are different, and I'm not sure if they will follow mores law. But it is very true and interesting how Moore's law follows for flash devices.

daveirl

This post has been deleted.

Capt'n Midnight

don't forget that most chips are 2D for esae of manufacture and cooling reasons. with some a better cooling system and better manu techniques you could make them 3D with many many layers so could gets lots of doublings that way too.

OfflerCrocGod

M's Law has nothing to do with Clock speed it's to do with trans density doubling every 12-18 months (originally it was 12 now it's 18 soon.....it could be changed to 24) It is not and never was a Natural Law it was more an observation of industry trends into the foreseeable future and then "Law" was stuck on to it, think feature size not clock speeds. Here comes 65nm still more to go.