Slashdot Mirror
45 Years Later, Does Moore's Law Still Hold True?
Velcroman1 writes "Intel has packed just shy of a billion transistors into the 216 square millimeters of silicon that compose its latest chip, each one far, far thinner than a sliver of human hair. But this mind-blowing feat of engineering doesn't really surprise us, right? After all, that's just Moore's Law in action isn't it? In 1965, an article in "Electronics" magazine by Gordon Moore, the future founder of chip juggernaut Intel, predicted that computer processing power would double roughly every 18 months. Or maybe he said 12 months. Or was it 24 months? Actually, nowhere in the article did Moore actually spell out that famous declaration, nor does the word 'law' even appear in the article at all. Yet the idea has proved remarkably resilient over time, entering the zeitgeist and lodging like a stubborn computer virus you just can't eradicate. But does it hold true? Strangely, that seems to depend more than anything on whom you ask. 'Yes, it still matters, and yes we're still tracking it,' said Mark Bohr, Intel senior fellow and director of process architecture and integration. 'Semiconductor chips haven't actually tracked the progress predicted by Moore's law for many years,' said Tom Halfhill, the well respected chip analyst with industry bible the Microprocessor Report."
Number of components, not computing power, and the time-frame should be easy to figure out given the difference between 1965's number and the 65,000 predicted in 1975.
Put identity in the browser.
Stop using Microsoft Windows.
45 Years Later, Does Moore's Law Still Matter?
Seriously, hardware is always getting faster. Why do we need a law that states this? Which is a more likely scenario for Intel: "Ok, we need to make our chips faster because of some ancient arbitrary rule of thumb for hardware speed.", or "Ok, we need to make our chips faster because if we don't, AMD will overtake us and we'll lose money."?
I'd say you'd run out of stories in which to post first, but it seems as though /. follows a Moore's law of duplicate stories so you should be alright.
Is there some corollary to Moore's law regarding the frequency at which articles will be written commemorating the age of Moore's law and asking if it is relevant?
Well, if it didn't, then would we still be talking about it, forty-five years later?
Am I part of the core demographic for Swedish Fish?
No, yes, no, no, no.
The real problem is access speeds. Even if you had an arbitrarily powerful CPU, you'd still have to load in everything from memory, hard disk, or network sources (i.e. all very slow). Until these can keep up the same pace as CPUs (SSDs are still expensive), it's pretty much just AMD and Intel having a pissing match. How often do you really max out your CPU cycles these days anyway?
Is it really that difficult?
The complexity for minimum component costs has increased at a rate of roughly a factor of two per year... Certainly over the short term this rate can be expected to continue, if not to increase. Over the longer term, the rate of increase is a bit more uncertain, although there is no reason to believe it will not remain nearly constant for at least 10 years. That means by 1975, the number of components per integrated circuit for minimum cost will be 65,000. I believe that such a large circuit can be built on a single wafer.[7]
Original Article:
Cramming more components
onto integrated circuit
Article 2: Excerpts from A Conversation
with Gordon Moore: Moore’s Law
So an article on Fox News concludes that it doesn't really matter if it's true so long as people think it's ok by consensus. Par for the course.
There's no question on what "Moore's Law" is as the article would paint. Originally, he said double transistor count every year. Then, in 1975, he revised it to every two years.
It's obviously not a scientific law but it is based on the manufacturing process for circuits and how they evolve, and it has been a good rule of thumb number and has proven as accurate as can be expected while we continue to make chips in basically the same ways.
It's fairly easy to look this up, there's no need for a lame mainstream media article link.
It's also been so frequently misused that Halfhill was forced to define Moron's Law, which states that "the number of ignorant references to Moore's Law doubles every 12 months."
There are only 13 posts so far, and yet /. is still on track to meet this law. Great job everyone.
Government is not reason; it is not eloquent; it is force. Like fire, it is a dangerous servant and a fearful master.
Well the problem here is that the question "Does Moore's Law Hold True?" is not very precise. It's easy to show both that the law doesn't hold, and that it is being followed still today, depending on how tight your definitions are.
If you extrapolate from the date that Moore first made the prediction, using the transistor counts of the day and a particular scaling exponent ("doubling every two years"), then the extrapolated line, today, will not exactly match current transistor counts. So it fails.
But if you use the "Law" in its most general form, which is something like "computing power will increase exponentially with time" then yes, it's basically true. One of the problems with this, however, is that you can draw a straight-line, and get a power-law exponent, through a lot of datasets once plotted in a log-linear fashion. To know whether the data "really is" following a power law, you need to do some more careful statistics, and decide on what you think the error bars are. Again, with sufficiently large error bars, our computing power is certainly increasing exponentially. But, on the other hand, if you do a careful fit you'll find the scaling law is not constant: it actually changes in different time periods (corresponding to breakthroughs and corresponding maturation of technology, for instance). So claiming that the history of computing fits a single exponent is an approximation, at best.
So you really need to be clear what question you're asking. If the question is asking whether "Moore's Law" is really an incontrovertible law, then the answer is "no". If the question is whether it's been a pretty good predictor, then answer is "yes" (depending on what you mean by "pretty good" of course). If the question is "Does industry still use some kind of assumption of exponential scaling in their roadmapping?" the answer is "yes" (just go look at the roadmaps). If the question is "Can this exponential scaling continue forever?" then the answer is "no" (there are fundamental limits to computation). If the question is "When will the microelectronics industry stop being able to deliver new computers with exponentially more power?" then the answer is "I don't know."
It's not a natural law. It's neither a law of physics nor one of biology. Heeding or ignoring it has no real meaning. And, bluntly, I doubt anyone but nerds that double as beancounters really care about Moore's "law".
Computers have to be fast enough to do what tasks they're supposed to solve. Software will grow to make use of it (or waste it on eye candy). Nobody but us cares about the rest.
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
What the fthagn is this? A Fox News article on /.? And it's actually accurate, non-politicized reporting on a scientific matter?
Apparently, I have entered the Bizarro World. Or perhaps the Mirror Universe. I can't be dreaming, because I'm not surrounded by hot women in tiny outfits, but something is most definitely WRONG here, and I aim to find out what.
At best it is a self-fulfilling prophesy, as the 'law' is now used as a standard for judging the industry, which strives to keep up with the predictions.
---- It won't be as bad as you fear or as good as you hope, but it will take twice as long as you plan.
I feel like I've been reading this article every six months for the last ten years.
...Moore's law is fucking stupid.
There, I said it.
In all seriousness, this is not like some sort of law of physics or something. It is just bloody stupid to keep quoting it all the time.
BTW, I want to add that I don't think Gordon Moore is stupid, only that the myth of this "law" is perpetuated.
Proverbs 21:19
The graph in Moore's article clearly predicts double the number of chips every 13 months. Nine "doublings" in 10 years.
...but I gave up caring about processor speed about 10 years ago.
I realize no one reads TFA; not even the submitter did. I don't know which is more absurd, the submitter claiming the above is not in TFA or someone ignoring TFA, reading the wiki page, then linking TFA as the source of Moore's Law without realizing it's TFA!
Head explodes
That is an amazingly shallow article on FoxNews.com Maybe they should advertise the source in the summary blurb. They seem to believe that computers will magically keep getting better without any specific hardware improvements required.
I clicked only wanting one thing, a graph with three lines showing: Moore's Law, transistor count, and computing power of each processor.
Even if you had an arbitrarily powerful CPU, you'd still have to load in everything from memory, hard disk, or network sources (i.e. all very slow)
Considering that light only travels 30 cm per nanosecond in a vacuum, the maximum practical clock speed depends on how far your memory is. At a 3 GHz clock rate, a request for data from a chip that's just 5 cm away on the circuit board will have a latency longer than the clock period.
The only solution to this problem is increasing the on-chip cache. But that will depend on having software that manages the cache well, i.e. more complex algorithms. In that case, since you have to optimize the software anyhow, why not go to a parallel architecture?
I bet that in the future we will see chips with simpler (read RISC) architectures with more on-chip memory and special compilers designed to optimize tasks to minimize random memory access.
First law of journalism.. never in your headline write a question that can easily be answered with a yes or a no....
Is it true that computing power has exactly doubled for 45 years over any specific length of time?... No... end of article.
Moore's law more than a technologcal vision, is a business strategy powered by a technological vision. I just see corporates pigs, (no ofence to pigs) doing unjustified capitalism. Now we should have other priorities.
None of these are 'laws', where you get punished by breaking them. Not Moore's, not Godwin's, etc. They are more 'generalizations' than anything else. Moore's, especially, could be more acccurately terms an 'observation', as that's what was going on at the time he made it. Everyone repeat after me: "Moore's Observation"
There we go.
Even "Moore's Average" would be more accurate.
the future founder of chip juggernaut Intel, predicted that computer processing power would double roughly every 18 months. Or maybe he said 12 months
What Gordon Moore actually said was that complexity would double every year. Moore was also relating cost at that time, but cost doesn't actually scale well, so most people don't include cost in modern interpretations of Moore's Law.
For circuit complexity, Moore's Law (with the 18 month amendment) seems to still hold true. However, we are fast approaching some physical limits that may cause the doubling period to increase.
Performance is commonly associated with Moore's Law (as you mention), However, performance is a function of clock speed, architecture, algorithm, and a host of other parameters and certainly does not follow Moore's Law... It never really has, even though people still like to think it does... or should...
Some people have thought Moore's law was soon to come to a close due to heat restrictions, but with the recent work on memsitors they believe chip technology will continue to grow steadily for quite sometime. I herd this from a professor of mine at the University of Alberta, take it for what you want.
CharlieMopps's Law(TM): The quantity of articles posted to Slashdot that mention Moore's Law will approximately double every time Intel or AMD come out with a new processor.
"Moore ... predicted that computer processing power would double roughly every 18 months. Or maybe he said 12 months. Or was it 24 months? Actually, nowhere in the article did Moore actually spell out that famous declaration, nor does the word "law" even appear in the article at all."
"The complexity for minimum component costs has increased at a rate of roughly a factor of two per year (see graph on next page). Certainly over the short term this rate can be expected to continue, if not to increase."
Moore's law is about minimum cost per unit. In traditional manufacturing, this is volume. In the IC world, component volume is simply chip density, but unit volume is affected adversely by chip density (complexity up, yields down). Thus, there is a balancing act. Moore's law is about the minimum, the optimal point on the curve from a manufacturer's standpoint.
This minimum has traditionally directly correlated with performance. Why? Because we tend to add components that end up doing work.
No, Moore didn't call it a law. It was an observation, and a prediction. As it proved to hold true over the next several decades, people referred to it as a law.
And it continues to hold true.
So please shut the fuck and stop asking about it every 6 months. Moore's law isn't about to be broken. Tablets aren't replacing desktop PCs any time soon. The cloud isn't changing the face of computing as we know it. This isn't the year of the Linux desktop. That breakthrough in solar panel efficiency will never materialize. Batteries still suck ass and there are no signs of any significant improvement on the horizon.
If you've got no news to report, just throw up "Not much news today. People are still talking about those new Intel CPUs, though.".
Take a look at top500.org and you can see a fairly steady doubling of performance every 13-14 months for the last 17 years or so. From what I understand, I'd expect the current growth trend to continue for the next few years.
Back in 1973 I made it to my first NCC (the AFIPS National Computer Conference - the annual big industry shindig in those days). At that time Moore's law was quite the buzz. Memory chips were still following it, but complex function chips were starting to fall off from the straight line on the log graph. At that time there were a few microprocessors out. But it was far before the stage where you could put a microprocessor on every device control card. Most such functions - including the "glue" around the microprocessors themselves - were constructed of small-scale integration chips. Support chips were starting to graduate from things like four independent gates, a couple flops, or a multiplexer per package. But chips were essentially all still being designed by silicon manufacturers. A few might have been done under contract with companies designing boxes. But most were based on the semiconductor companies' marketing departments' guess at what would be wanted a couple years in the future. I realized that one explanation for the shortfall might be that, as the complex function chips became larger, the engineering of more of the circuitry was moving from the system designers - including the garage and venture-financed startups - to the semiconductor manufacturers. This reduced the number of engineers on the job and their connection to the needs of the final products. Further, it changed the incentives on the engineers, making them more conservative (since they needed to keep an established company in business rather than take risks to establish a new venture or product). There was a panel with several of the silicon companies that discussed the problem. Come the Q and A session I brought up the above, and proposed a solution: That the silicon companies license their design tools to the system designers and build the chips THEY design. That way the complex-function engineering, along with its risks and costs, could be moved back to the ventures, while the silicon companies could concentrate their engineering on what they do well - improving the process. And I asked whether any of their companies would consider such an approach. (I thought of it as a "silicon breadboard", but I don't recall actually using the term in the question.) At least three of the companies' representatives - Mororola, Intel, I forget who else - said that there was no way they would ever do such a thing. (The Motorola guy was quite emphatic about it.) And the guy beside me gave me his card and suggested I interview with him. (He was from Signetics, which was already doing a mask-programmed gate array chip which the customer could customize. I DID interview with him - and to this day I kick myself for not taking a job there. It would have gotten me out to Silicon Valley 12 years earlier, two years before both the release of the Altair 8080 and the founding of the Homebrew Computer Club. B-b ) A few months later that year, IBM announced they'd make their design tools available to customers and would fabricate chips under contract. Over the next couple years several other manufacturers followed suit. One of them transitioned from custom silicon design to tool licensing as a business and several others started up just to do tools. For a while it was known as the "silicon foundry" system. Now it's ASIC (application specific integrated circuit) design, there are standards for the major design languages, and a whole ecosystem of manufacturers of chips and of computer-aided design tools for all stages of the process. And ASIC design is what I do for a living since I went back over to the hard side of the force in the early 1990s.
What's up with the inane mention of Pirates of the Carribean? The writer even got it wrong, as the first one to utter the words was Geoffrey Rush's character (Ctp. Barbossa). It's probably a FauxNews specialty...
(Trying again with paragraph breaks. B-b )
Back in 1973 I made it to my first NCC (the AFIPS National Computer Conference - the annual big industry shindig in those days). At that time Moore's law was quite the buzz. Memory chips were still following it, but complex function chips were starting to fall off from the straight line on the log graph.
At that time there were a few microprocessors out. But it was far before the stage where you could put a microprocessor on every device control card. Most such functions - including the "glue" around the microprocessors themselves - were constructed of small-scale integration chips. Support chips were starting to graduate from things like four independent gates, a couple flops, or a multiplexer per package. But chips were essentially all still being designed by silicon manufacturers. A few might have been done under contract with companies designing boxes. But most were based on the semiconductor companies' marketing departments' guess at what would be wanted a couple years in the future.
I realized that one explanation for the shortfall might be that, as the complex function chips became larger, the engineering of more of the circuitry was moving from the system designers - including the garage and venture-financed startups - to the semiconductor manufacturers. This reduced the number of engineers on the job and their connection to the needs of the final products. Further, it changed the incentives on the engineers, making them more conservative (since they needed to keep an established company in business rather than take risks to establish a new venture or product).
There was a panel with several of the silicon companies that discussed the problem. Come the Q and A session I brought up the above, and proposed a solution: That the silicon companies license their design tools to the system designers and build the chips THEY design. That way the complex-function engineering, along with its risks and costs, could be moved back to the ventures, while the silicon companies could concentrate their engineering on what they do well - improving the process. And I asked whether any of their companies would consider such an approach. (I thought of it as a "silicon breadboard", but I don't recall actually using the term in the question.)
At least three of the companies' representatives - Mororola, Intel, I forget who else - said that there was no way they would ever do such a thing. (The Motorola guy was quite emphatic about it.)
And the guy beside me gave me his card and suggested I interview with him. (He was from Signetics, which was already doing a mask-programmed gate array chip which the customer could customize. I DID interview with him - and to this day I kick myself for not taking a job there. It would have gotten me out to Silicon Valley 12 years earlier, two years before both the release of the Altair 8080 and the founding of the Homebrew Computer Club. B-b )
A few months later that year, IBM announced they'd make their design tools available to customers and would fabricate chips under contract. Over the next couple years several other manufacturers followed suit. One of them transitioned from custom silicon design to tool licensing as a business and several others started up just to do tools. For a while it was known as the "silicon foundry" system. Now it's ASIC (application specific integrated circuit) design, there are standards for the major design languages, and a whole ecosystem of manufacturers of chips and of computer-aided design tools for all stages of the process.
And ASIC design is what I do for a living since I went back over to the hard side of the force in the early 1990s.
I'd say that the fact that computing PRODUCTS have largely tracked "Moore's Law" says more about market forces and competition, and "Wintel" (Microsoft/bloat/software purchases, etc), than it says about physics, engineering and computing technology... It says more about what kind of products and features are needed to drive the IT money machine to spend and spend even though actually the computing needs to write letters, emails and most documents was attained more than a decade ago. Don't forget about advertising in the equation, all those Flash-drive websites throwing ads at you, requiring 2-4GB of RAM and plenty of CPU and GPU just to support a browser to display all that marketing crap without a true increase in substantive content...
Technology often follows an "ess-curve", a steep upward slope of exponential improvement followed by a flattening, diminished returns. Micro-electronics is still in the exponential growth part, but will flatten some unknown decade hence. Its always been a decade or two in the future in my life. Trains and automobiles are technologies that achieved most of their efficiencies now. Although the incorporation of computing has reinvigorated them somewhat.
Moore wasn't the first person to recognize this, only the most famous. See for example: http://docs.google.com/viewer?a=v&pid=explorer&chrome=true&srcid=1PX1HUsIBhnOSOsxcM2HoJZHv-9rtnNPOu-CA--584-UA4GLXxhWz842JmRDy&hl=en which describes Selengut's results from 1959. Selengut predicted computer speeds would double in magnitude every 10 years.
1) A law does not imply causality.
2) Moore's Law does not state that processing power doubles every 2 years. It states that the number of transistors that can be placed reasonably economically on an integrated circuit doubles every 2 years. It's not the same thing.
...had more to do with the *cost* of transistors than the quantity or speed of them. The increasing quantities per device are, however, simultaneously a cause (economy of scale) and effect (your dollar goes further) of this reducing cost.
We can now buy a device with a billion transistors in it for a couple of hundred dollars. That's about 50K transistors per cent. It follows that about 30 years ago, it would have been about 1 transistor per cent, or in other words the original 68000 would have cost about $680 in about 1980. Since it is described as originally being "quite expensive", this seems to fit. I don't have any data on the cost of transistors in 1965, though.
--- The key to knowledge is not to rely on people to teach you it ---
I feel like I see twice the rate of stories about this every 18 months or so.
Look at PCs and storage as the bellwether here - both have considerably declined in price per unit of (whatever) over time. Storage was $1 per byte for RAM around the time Moore made his prognostication. Disks held 5MB per 14" platter. Today quite a lot more in quite a lot less volume. Same for compute power. A million transistors on a chip seemed like a lot just 20 years ago. Now we're pushing a billion and the CPUs still cost less than they did a score ago. Is it an exact straight line over time? Of course not. But I wish I could get the same value per dollar from just about any other commodity. Pizzas would be about .001 cents per slice by now. Mmmmm Pizzas.
Chip makers intentionally regulate (slow down) their advancement to meet Moore's law because it allows them to make greater profits by forcing user's to upgrade on a regular basis, while still giving them enough time to thoroughly test the next iteration and make a profit on it.
Sure, I don't care why intel is making their chips faster, but I would like to know how much faster and how?
If you have a software project, scheduled for three years of development, can I rely on my average customer to be running computers 2 ^ 1.5 times as fast as they are now, or will multi-core machines proliferate?
As an intel share holder, all I'd need is your question, but as a computer user looking to the future, I'm more interested in the answer to the original question.
http://en.wikipedia.org/wiki/Moore's_law
That is because you have, and it is even less relevant now. I mean so what if we haven't hit the wall on Moore's Law yet, we sure as hell hit it when it came to speed VS heat. So now everyone is having to trade more cores for raw speed only problem? It does not actually work beyond a couple of cores for 95%+ of the population. They just don't have enough jobs for the cores to do and most jobs are still depending on single threaded performance.
So they can reprint this article every year like Moore's Law is the gospel of Gordon or something but to the average folks it just doesn't matter. Trying to increase performance by constantly adding more cores (what are we up to, 6 each for AMD and Intel?) is putting band aids on the bullet wound. Most jobs simply don't benefit from more cores, and we can't figure out how to fix the heat problem above 3.6GHz.
It is a shame we are down to just two CPU manufacturers, because if someone could find a way to make a 5GHz+ CPU triple core that doesn't get so hot you could cook your lunch on it? Well THAT would be a real breakthrough. But at the way we are going we are gonna end up with everyone and their grandma on an octacore CPU that spends most of its time twiddling its thumbs.
ACs don't waste your time replying, your posts are never seen by me.
Fabricated on 25nm: http://www.intel.com/pressroom/archive/releases/2010/20100201comp.htm over 36 billion transistors in 167 mm^2.
FusionGyro's Constant
Upon tracking various technology articles over the last decade, FusionGyro's observed semi-annual publishing of article's questioning whether Moore's Observation would still hold true
You don't belong here if you RTFA!
The US government have made it clear that we have no inalienable rights; any we do not defend vigorously will be taken.
Unfortunately its the aggregate of a pile of small independent undemanding tasks that drags modern PCs to a crawl. And these aren't even bottlenecking the CPU itself... to be honest I don't know what the bottleneck is right now in some items... I'll open up the task manager... cpu utilization will be comfortably low on all cores, hard drive lights are idle so it shouldn't be waiting on IO... and the progress bar is just sitting there... literally 20-30 seconds later things start happening again... WHAT THE HELL? What are the possible bottlenecks that cause this?
Might be thrashing or a task scheduler issue. Try a different OS on that machine to see if the performance changes, or try a similar machine configured with more memory.
But it might also be the Von Neumann Bottleneck. If your working set - either data or instructions - is bigger than the on-chip caches, you get constant cache misses. It's like thrashing but between cache and memory rather than between memory and the swap partition.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Seriously I remember hearing it back in the 70's. Enough already. Transistors are so small that human hair isn't even a reasonable comparison to make.
It was Geoffrey Rush as Captain Barbossa and not Johnny Depp who first used the phrase "guidelines" in the Pirates movies.
Is this enough of an error to doubt the veracity of the rest of the article? ;-)
My other Slashdot ID is much lower.