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.
as an excuse for a lack of innovation?
"we have reached the limits so don't expect innovation!"
...and therein lies its true flaw. As the law stipulates doubling transistor counts, as soon as processors are primarily developed with non-transistor based technologies, be they optical or quantum derived, Moore's Law is essentially defunct.
Well, it's impressive for a "law" which is not in any fundamental sense a law, but a speculation about future progress.
Very few speculations hold for so long.
By the way, I assume your account name is pronounced "fish".
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
Perhaps Moore's law really is beginning to run up against its limits, as you will see if you read enough electronics magazines, but what I really don't "get" is this: The Intel processor can do amazing things, but look at the Motorola processors, like the G4s in those Macs... They're faster at floating point and at a variety of other uses. Their instruction set is quite different. There are many other significant differences between the Intel and Motorola processors. And as we know from software, the way an algorithm is made up, or the way it is implemented, can drastically affect the performance. I think processors follow quite the same rules. Maybe it's time, while we're running up against the limits of Moore's law, to examine what software needs to do nowadays, and then design a processor from the ground up that will fulfill each function in the most efficient way possible. And while we're at it, let's go back to the good ol' days of making the software efficient, too. You'd be amazed the kinds of ridiculous things todays' computers can do, but the software is just too darn inefficient.
Why is it that every post that says "i will probably get modded down" actually gets modded up?
(mod me down if you want)
Quantum computing is neat in theory, but has made not significant progress in the number of qbits manipulatable in years. Granted there are new ways to make qbits, but nothing can seem to get 7 to 10 to date. Hopefully there will be a breakthrough, but you can't just command one. There is no scaling technology for Quantum Computers yet.
I predict biological approaches will similarly run into intractably hard roadblocks on the way to usefulness, with the possible exception of practical biological to electronic interfaces to aid the disabled and in the more distant future meld with the machine so to speak.
All is not lost however, multicore is of course where the industry is going for now, but expect more specialization in silicon for well-defined tasks. Graphics processors will get more powerful as algorithms improve and are more efficiently implemented with the transistors available. Any application that becomes mainstream will get its own processing unit of some sort. Granted this make for less flexibility in expanding the capabilities of existing machines, but software has been getting a free ride off the speed scaling in chips for years. In the future the line between programming and chip designing will blur as the two must work in concert to achieve the desired performance in whatever domain is desired.
Imagine a compiler that doesn't just compile code but tapes out the coprocessor need to run it.
Letter To Iran