Chris, I'm using the same one that Moore uses, and the same exact one that the article I linked describes: The number of transistors on a chip doubles roughly every 18-24 months.
This is not the only thing in the IT field that is still on a decades-long exponential growth curve of course, but different parts are moving at different rates. Storage, clock ratings, MIPS, etc. all also tend to double, but some more quickly and some more slowly than an 18-24 month cycle.
The sloppiness of the term came in when people started using "Moore's Law" to describe the exponential growth of tech in general. Which works as a rule of thumb but is far more generalized than anything Moore ever said.
I suggest reading this article if you want to really be informed about the discussion.
Only since the 1990s, Dean? I first heard we were hitting the limit with bubble memory back in 1978.
Moore's Law in the specific case leads to the sloppy but commonly accepted Moore's Law in the general case. Those twice as many transistors will be put to use doing new things that just weren't practical before, and will inspire new things that weren't even imagined before. The range of cool tools and applications we'll be using five years from now is almost literally beyond our comprehension today.
I think maybe you're a little older Martin. In any case I hadn't even heard of Moore's Law until the late 1980s, so I didn't hear that it was hitting the wall until the early 1990s.
Hah. Only a couple years older, Dean. I just was more of a geek. When I heard this, it was at a deep philosophical conversation at my first Star Trek convention. The guy I heard it from was a very serious geek, despite the fact that he had a miniature of the monster from Alien sitting on his shoulder.
And come to think if it, it had to be 1979, not 1978. I drove to that convention, so I had my license, and it was late fall. That would make it 1979.
I got into computers right around the time when the price of blazingly fast 450 nanosecond ram plummeted from over $100 for 16K to under $20. A little more recently than 1978, and I remember discussion then as to how much more dense semiconductors could become.
There was also discussion, parallel, as you say, to Moore's law as to how dense magnetic recording on a disk could get. I scavenged an early SGI workstation which contained a 158 megabyte hard drive - 5-1/4 inch full height - which had to have been state of the art when ordered.
By the standards of 1980, when I soldered together my first computer, I have a million dollars worth of computer capacity sitting idle under my workbench.
I have to admit that I skimmed the article, so I didn't see where they defined which version of Moore's law they were talking about. I don't think that you can actually claim that there's an authoritative version of the "law", though, since Moore himself never meant it as a law. He made an observation about a while ago, and people turned it into a "law" colloquially.
Also, a "law" with uncertainty in it (18-24?) isn't much of a law. Imagine if the law of gravity was F=Gm1m1/(d^2 ish).
What's really bad about the generalization is that nothing else doubles every 18-24 months. Cars don't go twice as fast, or double in fuel efficiency, or carrying capacity, or safety, or in fact in anything at all in 18-24 months. Ditto for airplanes, light bulbs, building materials (when's the last time that steel doubled in strength?), land-line telephones, etc. Monitors don't double in resolution, or size, or refresh rate, or brightness, or anything. In fact, I don't think that you can find any other technologies which have consistently doubled (or halved) in anything every two years consistently over two decades.
Chris: So you only skimmed the article but you want to comment on it anyway?
Moore specifically stated that the number of transistors doubles every 18-24 months. It became a law when it became obvious that this was happening consistently over decades, and it's still going on.
Also, with respect, you really can't understand math very well if you think that statistical variability cannot be part of any scientific law. Also, you're mistaken if you believe that this is the only area of technology that's doubling every few years; you are confusing technologies that have reached useful maturity with all technology. I would recommend doing some more serious reading and thinking about this before continuing to so cavalierly dismiss it; we're seeing exponential growth in a vast number of subjects, over several important fields.
I wasn't saying that the thing was wrong; that would be asinine to do after skimming it. But clarifying what you mean is hardly illegitimate after not thoroughly reading someone else. The problem is that over the years either you or people who closely agree with you (and who you never corrected) have used the term "Moore's Law" about more than just transistor counts.
In short, I was more interested in what you got out of the article than what's actually in it. (Incidentally, I should ask if you're talking about Moore's Law including on-chip caches in the transistor count, or excluding caches, because my recollection is that you get different validity for this "law" if you exclude caches.)
And Dean, if you really think that wishy washy fudge factors are part of "laws", you're the one who needs to re-think things. Especially in mathematics, there are precisely no laws in which the statement is vague and uncertain. There are certainly laws having to do with probabilities, and laws having to do with ranges, but none of them simply fudge the claim. But the problem is that "law" is not well defined, since we never clarified what sort of law. This is certainly not a mathematical law, at least not in its ordinary statement.
And the truth is, it's not any other sort of law, either. It can't continue forever, if for no other reason than that there are a finite number of particles in the universe, and you can't make infinitely many transistors out of finitely man particles.
Look, name me another technology which has consistently doubled every 18-24 months for two decades. Of course new fields experience exponential growth, that's not at all the same thing as keeping up the exponential growth for 20 generations of doubling. Nor for forty generations of doubling.
But that's what moore's law means: that transistors will never be a mature technology.
Look, the reason that this bugs me so much is that the way you talk about processors and transistors is precisely the way that people talk about the stock market during the bubbles which precede crashes. You don't talk about underlying mechanisms, you only talk about past performance. It's unsettling because that type of reasoning has lead to huge disasters in the past.
(But about the particulars, I dearly hope that you're right. Processors aren't nearly fast enough, and neither is RAM or disk space. Actually, there's another one that has doubled an amazing amount. And I believe that that will keep doubling for quite some time too. But RAM hasn't been doubling in size. There was a very long time there where RAM wasn't getting any bigger (at the same price), it was only getting faster. Hopefully vista will finally push things in the direction of getting bigger again.)
Dr. Moore himself is reluctant to call it a law. That's just a name that has stuck.
On the other hand, while it's clearly true that past performance does not guarantee future results, I do like to scoff at the people screaming "Limit!" because so many have screamed it so many times in the past. While I know with absolute certainty that there must be a limit, I've come to the conclusion that anyone who predicts when it will hit is wrong -- and the sooner they predict it, the more laughably wrong they will be. I think we'll only know what the limit is after we hit it.
So in short, Chris: you laugh and say "this can't go on forever." And I say, "Yes, but I've been listening to that particular line of argument for almost as long as you've been alive and it hasn't happened yet."
My only argument for you otherwise is that you're simply wrong: scientific laws often DO have fudging--otherwise known as "averages"--around the edges.
Indeed, the entirety of Darwinism rests upon just such averages. Is there a Law of Natural Selection? Or is it simply a fact that evolution continues apace all around us?
It may be that Moore's Law should not be stated as a law. Probably that's true. Nevertheless that's the appellation it's currently known by, and so far it has continued to remain true. After nearly 20 years of listening to how it's a joke and won't hold true much longer (or has already stopped), and watching it continue to hold true even while people assure me that it ain't so, it's hard for me not to laugh.
When you give me answers like "cars haven't continued to double in speed in decades" all I can think is, "wow, you haven't really read much about this, have you?" Cars some time ago reached a certain--probably temporary--technological maturity when they reached speeds beyond which humans could safely control them. We can and do make cars that are massively faster than your average production vehicle. We can do it right now. In fact we've built cars that can break the sound barrier. The only limit--the limit that has caused a maturity in the field for the moment--is that no human being can safely control a car that goes over 70 or 80 miles per hour for any sustained period of time. It's not that we can't go faster, it's just that it's not safe anymore.
Until the cars drive themselves, anyway.
The only other thing I'd observe for you is what I've said before: you're foolish if you think that exponential technological growth is restricted solely to the number of transistors on a chip. It's happening in multiple areas, not just this one.
I'd suggest doing more than skimming on this matter. It has profound implications, after all, and is not mere utopianism. Indeed, dismissing it all as utopianism is terribly seductive; if you dismiss it all as utopianism, you automatically lock yourself out of very important conversations that need to take place between intelligent people.
1.29.2007 9:24pm
Commenting on Dean's World is a privilege, not a right. Dean is your host, you are his guest, and you should behave in that fashion. Dean is not your babysitter, nor is he your punching bag. Please remember this. In general, you are free to disagree with anyone on any subject you wish, but abusive behavior will not be tolerated.
Of course we all lose our tempers now and then. Dean freely admits to being imperfect in this regard, which is why regulars to this establishment will generally be cut more slack than people who we don't know very well.
Still: behave like an adult, or go find somewhere else to play. Thanks.
This is not the only thing in the IT field that is still on a decades-long exponential growth curve of course, but different parts are moving at different rates. Storage, clock ratings, MIPS, etc. all also tend to double, but some more quickly and some more slowly than an 18-24 month cycle.
The sloppiness of the term came in when people started using "Moore's Law" to describe the exponential growth of tech in general. Which works as a rule of thumb but is far more generalized than anything Moore ever said.
I suggest reading this article if you want to really be informed about the discussion.
Moore's Law in the specific case leads to the sloppy but commonly accepted Moore's Law in the general case. Those twice as many transistors will be put to use doing new things that just weren't practical before, and will inspire new things that weren't even imagined before. The range of cool tools and applications we'll be using five years from now is almost literally beyond our comprehension today.
You're otherwise right I suspect.
And come to think if it, it had to be 1979, not 1978. I drove to that convention, so I had my license, and it was late fall. That would make it 1979.
There was also discussion, parallel, as you say, to Moore's law as to how dense magnetic recording on a disk could get. I scavenged an early SGI workstation which contained a 158 megabyte hard drive - 5-1/4 inch full height - which had to have been state of the art when ordered.
By the standards of 1980, when I soldered together my first computer, I have a million dollars worth of computer capacity sitting idle under my workbench.
I have to admit that I skimmed the article, so I didn't see where they defined which version of Moore's law they were talking about. I don't think that you can actually claim that there's an authoritative version of the "law", though, since Moore himself never meant it as a law. He made an observation about a while ago, and people turned it into a "law" colloquially.
Also, a "law" with uncertainty in it (18-24?) isn't much of a law. Imagine if the law of gravity was F=Gm1m1/(d^2 ish).
What's really bad about the generalization is that nothing else doubles every 18-24 months. Cars don't go twice as fast, or double in fuel efficiency, or carrying capacity, or safety, or in fact in anything at all in 18-24 months. Ditto for airplanes, light bulbs, building materials (when's the last time that steel doubled in strength?), land-line telephones, etc. Monitors don't double in resolution, or size, or refresh rate, or brightness, or anything. In fact, I don't think that you can find any other technologies which have consistently doubled (or halved) in anything every two years consistently over two decades.
Hahahaha.
Sorry but that was funny.
Moore specifically stated that the number of transistors doubles every 18-24 months. It became a law when it became obvious that this was happening consistently over decades, and it's still going on.
Also, with respect, you really can't understand math very well if you think that statistical variability cannot be part of any scientific law. Also, you're mistaken if you believe that this is the only area of technology that's doubling every few years; you are confusing technologies that have reached useful maturity with all technology. I would recommend doing some more serious reading and thinking about this before continuing to so cavalierly dismiss it; we're seeing exponential growth in a vast number of subjects, over several important fields.
I wasn't saying that the thing was wrong; that would be asinine to do after skimming it. But clarifying what you mean is hardly illegitimate after not thoroughly reading someone else. The problem is that over the years either you or people who closely agree with you (and who you never corrected) have used the term "Moore's Law" about more than just transistor counts.
In short, I was more interested in what you got out of the article than what's actually in it. (Incidentally, I should ask if you're talking about Moore's Law including on-chip caches in the transistor count, or excluding caches, because my recollection is that you get different validity for this "law" if you exclude caches.)
And Dean, if you really think that wishy washy fudge factors are part of "laws", you're the one who needs to re-think things. Especially in mathematics, there are precisely no laws in which the statement is vague and uncertain. There are certainly laws having to do with probabilities, and laws having to do with ranges, but none of them simply fudge the claim. But the problem is that "law" is not well defined, since we never clarified what sort of law. This is certainly not a mathematical law, at least not in its ordinary statement.
And the truth is, it's not any other sort of law, either. It can't continue forever, if for no other reason than that there are a finite number of particles in the universe, and you can't make infinitely many transistors out of finitely man particles.
Look, name me another technology which has consistently doubled every 18-24 months for two decades. Of course new fields experience exponential growth, that's not at all the same thing as keeping up the exponential growth for 20 generations of doubling. Nor for forty generations of doubling.
But that's what moore's law means: that transistors will never be a mature technology.
Look, the reason that this bugs me so much is that the way you talk about processors and transistors is precisely the way that people talk about the stock market during the bubbles which precede crashes. You don't talk about underlying mechanisms, you only talk about past performance. It's unsettling because that type of reasoning has lead to huge disasters in the past.
(But about the particulars, I dearly hope that you're right. Processors aren't nearly fast enough, and neither is RAM or disk space. Actually, there's another one that has doubled an amazing amount. And I believe that that will keep doubling for quite some time too. But RAM hasn't been doubling in size. There was a very long time there where RAM wasn't getting any bigger (at the same price), it was only getting faster. Hopefully vista will finally push things in the direction of getting bigger again.)
Dr. Moore himself is reluctant to call it a law. That's just a name that has stuck.
On the other hand, while it's clearly true that past performance does not guarantee future results, I do like to scoff at the people screaming "Limit!" because so many have screamed it so many times in the past. While I know with absolute certainty that there must be a limit, I've come to the conclusion that anyone who predicts when it will hit is wrong -- and the sooner they predict it, the more laughably wrong they will be. I think we'll only know what the limit is after we hit it.
My only argument for you otherwise is that you're simply wrong: scientific laws often DO have fudging--otherwise known as "averages"--around the edges.
Indeed, the entirety of Darwinism rests upon just such averages. Is there a Law of Natural Selection? Or is it simply a fact that evolution continues apace all around us?
It may be that Moore's Law should not be stated as a law. Probably that's true. Nevertheless that's the appellation it's currently known by, and so far it has continued to remain true. After nearly 20 years of listening to how it's a joke and won't hold true much longer (or has already stopped), and watching it continue to hold true even while people assure me that it ain't so, it's hard for me not to laugh.
When you give me answers like "cars haven't continued to double in speed in decades" all I can think is, "wow, you haven't really read much about this, have you?" Cars some time ago reached a certain--probably temporary--technological maturity when they reached speeds beyond which humans could safely control them. We can and do make cars that are massively faster than your average production vehicle. We can do it right now. In fact we've built cars that can break the sound barrier. The only limit--the limit that has caused a maturity in the field for the moment--is that no human being can safely control a car that goes over 70 or 80 miles per hour for any sustained period of time. It's not that we can't go faster, it's just that it's not safe anymore.
Until the cars drive themselves, anyway.
The only other thing I'd observe for you is what I've said before: you're foolish if you think that exponential technological growth is restricted solely to the number of transistors on a chip. It's happening in multiple areas, not just this one.
I'd suggest doing more than skimming on this matter. It has profound implications, after all, and is not mere utopianism. Indeed, dismissing it all as utopianism is terribly seductive; if you dismiss it all as utopianism, you automatically lock yourself out of very important conversations that need to take place between intelligent people.
Of course we all lose our tempers now and then. Dean freely admits to being imperfect in this regard, which is why regulars to this establishment will generally be cut more slack than people who we don't know very well.
Still: behave like an adult, or go find somewhere else to play. Thanks.