Slashdot Mirror
Fifty Years of Moore's Law
HughPickens.com writes: IEEE is running a special report on "50 Years of Moore's Law" that considers "the gift that keeps on giving" from different points of view. Chris Mack begins by arguing that nothing about Moore's Law was inevitable. "Instead, it's a testament to hard work, human ingenuity, and the incentives of a free market. Moore's prediction may have started out as a fairly simple observation of a young industry. But over time it became an expectation and self-fulfilling prophecy—an ongoing act of creation by engineers and companies that saw the benefits of Moore's Law and did their best to keep it going, or else risk falling behind the competition."
Andrew "bunnie" Huang argues that Moore's Law is slowing and will someday stop, but the death of Moore's Law will spur innovation. "Someday in the foreseeable future, you will not be able to buy a better computer next year," writes Huang. "Under such a regime, you'll probably want to purchase things that are more nicely made to begin with. The idea of an "heirloom laptop" may sound preposterous today, but someday we may perceive our computers as cherished and useful looms to hand down to our children, much as some people today regard wristwatches or antique furniture."
Vaclav Smil writes about "Moore's Curse" and argues that there is a dark side to the revolution in electronics for it has had the unintended effect of raising expectations for technical progress. "We are assured that rapid progress will soon bring self-driving electric cars, hypersonic airplanes, individually tailored cancer cures, and instant three-dimensional printing of hearts and kidneys. We are even told it will pave the world's transition from fossil fuels to renewable energies," writes Smil. "But the doubling time for transistor density is no guide to technical progress generally. Modern life depends on many processes that improve rather slowly, not least the production of food and energy and the transportation of people and goods."
Finally, Cyrus Mody tackles the question: what kind of thing is Moore's Law? "Moore's Law is a human construct. As with legislation, though, most of us have little and only indirect say in its construction," writes Mody. "Everyone, both the producers and consumers of microelectronics, takes steps needed to maintain Moore's Law, yet everyone's experience is that they are subject to it."
Andrew "bunnie" Huang argues that Moore's Law is slowing and will someday stop, but the death of Moore's Law will spur innovation. "Someday in the foreseeable future, you will not be able to buy a better computer next year," writes Huang. "Under such a regime, you'll probably want to purchase things that are more nicely made to begin with. The idea of an "heirloom laptop" may sound preposterous today, but someday we may perceive our computers as cherished and useful looms to hand down to our children, much as some people today regard wristwatches or antique furniture."
Vaclav Smil writes about "Moore's Curse" and argues that there is a dark side to the revolution in electronics for it has had the unintended effect of raising expectations for technical progress. "We are assured that rapid progress will soon bring self-driving electric cars, hypersonic airplanes, individually tailored cancer cures, and instant three-dimensional printing of hearts and kidneys. We are even told it will pave the world's transition from fossil fuels to renewable energies," writes Smil. "But the doubling time for transistor density is no guide to technical progress generally. Modern life depends on many processes that improve rather slowly, not least the production of food and energy and the transportation of people and goods."
Finally, Cyrus Mody tackles the question: what kind of thing is Moore's Law? "Moore's Law is a human construct. As with legislation, though, most of us have little and only indirect say in its construction," writes Mody. "Everyone, both the producers and consumers of microelectronics, takes steps needed to maintain Moore's Law, yet everyone's experience is that they are subject to it."
> Andrew "bunnie" Huang argues that Moore's Law is slowing and will someday stop,
I think we've been hearing about the end of Moore's law for the last 15 years... inevitably, some process improvement comes along and it all keeps on going.
Yeah, it may "eventually" stop when transistors are built with just 3 atoms. Then will switch over to photonics or quantum, then some weird hyper-dimensional shit.
That guy is going to be pissed when we don't get cold supercomputers with billions of times more power than the brain using reversible computing.
Uh....
Well...
cough-cough
You see...
Aww screw it.
Could there have been worse examples of "LOL those crazy promises!"?
Dewey, what part of this looks like authorities should be involved?
The idea of an "heirloom laptop" may sound preposterous today, but someday we may perceive our computers as cherished and useful looms to hand down to our children, much as some people today regard wristwatches or antique furniture."
It is preposterous... Even if it were impossible to make computers faster in any way in the future (extremely unlikely given the countless avenues there are to explore in terms of speed), even then the inovation in computers i not and would not be limited to speed, so no computer heirlooms wont ever happen, stupid person.
Incidentally Moore's law died sometime last year technically, as Intel failed to ship its new node within "18-24 months" of its last one, meaning the density of transistors did not, for anyone, double within the time limits specified by Moore's Law. With the other foundries (TSMC/GloFlo/Samsung) still ramping up the same feature density size with finfet transistors that Intel had 3 years ago, and 10nm bringing even more difficulties than Intel's "14nm" it's a question how much longer feature size can continue to shrink at all, let alone somehow coming within the Moore's Law cadence of ever 18-24 months.
That "weird hyperdimensional shut" is the kind of innovation he is talking about. That stuff doesn't just fall from the sky. Lots of people have to innovate the he'll out of that stuff for years or decades. What website do you think you are reading, anyway?
Andrew "bunnie" Huang argues that Moore's Law is slowing and will someday stop
Moore's Meta-Law:
The number of people predicting the end of Moore's Law doubles every eighteen months!
I remember watching Star Trek (TOS) and thinking how fantastic it would be to have all that storage in that little cartridge the size of a matchbook; books, movies, medical records, the Encyclopedia Galactica, all on one little memory device. I never expected it happen in my lifetime.
Then in 1985 once the initial glow of the original Macintosh had worn off a little, my brother and I brainstormed on what our _ultimate_ computer would be: 1024x768 TrueColor display, a whole _8_ megabytes of memory, and a 50 Mhz 68000 series CPU. Wheee!
Now we have 128 GB microSD cards smaller than your fingernail. And that super-computer in your pocket that happens to make phone calls? It's more powerful than a 4 processor Cray YMP M90 circa 1992.
We've come a long way!
--aj;
I think we've been hearing about the end of Moore's law for the last 15 years... inevitably, some process improvement comes along and it all keeps on going.
I don't think that it's necessarily "inevitable". Take aviation, for example. There was arguably exponential increases in the capability of aircraft for 55 years from 1903 to 1958, when the Boeing 707 was introduced. Ever since, further progress on economically viable aircraft has been pretty much limited to incremental increases in fuel economy and marketing strategies to keep costs down by keeping planes full.
Moore's law is sort of a mangled version of Koomey's law. Koomey's law states that the number of computations per joule of energy dissipated has been doubling every 1.6 years. It appears to have been operative since the late 1940s: longer than Moore's law. Moreover, Koomey's law has the appeal of being defined in terms of basic physics, rather than technological artefacts. Hence, I prefer Koomey's law, even though Moore's law is far more famous.
There is another interesting aspect to Koomey's law: it hints at an answer to the question "for how long can this continue?" The hinted answer is "until 2050", because by 2050 computations will require so little energy that they will face a fundamental thermodynamic constraint—Landauer's principle. The only way to avoid that constraint is with reversible computing.
... under a loose interpretation. Mooers' law is 56.
Regarding Andrew âoebunnieâ Huang ridiculous article....
As commercial success and product differentiation starts to depend less on quickly leveraging the latest hardware and moreso on algorithmic improvements, companies will not magically become more inclined to publish source code. When the path to improved performance involves massive man-hours optimizing code, small teams & startups will not somehow gain an advantage.
Click baiting "open source" and an interactive graph might bring a lot of page views, but the entire premise is truly absurd.
PJRC: Electronic Projects, 8051 Microcontroller Tools
I am feeling sick and sad that my generation could be the failure that couldn't keep up with Moore's law and is looking for excuses and marketing incompetence as innovation.
Specially considering that we can't even fucking go to the moon anymore, and the motherfuckers who did it used fucking 64kb computers.
2 out of 2. We are self-appointed lazy losers full of ourselves and deserve no respect from our ancestors.
Chris Mack begins by arguing that nothing about Moore's Law was inevitable. "Instead, it's a testament to hard work, human ingenuity, and the incentives of a free market.
Humans working hard and having ingenuity, and being incentivized by the free market are all things that are sort of inevitable in themselves. I don't mean to diminish those positive features of humanity, but I think it's ok to take them for granted in the sense that I don't think it is likely for those things to stop being features of humanity barring some kind of catastrophe.
Was Moore's Law going to be as true as it was with 100% probability? No, some stuff could have gone wrong. Some people might have decided not to work so hard for whatever reason. But if we could rewind the clock and do it again, I think there is a very good chance it would turn out pretty much the same way, because I would expect people to be just as enthusiastic about making better and better solutions to problems. It's who we are.
In this respect, I think Moore's law was inevitable in the sense that another outcome like "win the lottery" isn't. We are a species that has had hundreds of thousands of years of evolutionary training in creative problem solving. Not only was maintaining Moore's law for this long within our capability to figure out, we were driven to do it for the same reasons we were driven to eat food and procreate.
Was it inevitable for a particular monkey to swing through the trees? It's true that he/she could have decided not to, but the monkey was destined to swing through trees in a way that a fish was not.
From the summary:
But the doubling time for transistor density is no guide to technical progress generally. Modern life depends on many processes that improve rather slowly, not least the production of food and energy and the transportation of people and goods.
A lot of progress depends on information technology, though. For example our understanding of biochemical processes. Or the capability of satellites that monitor what's going on with our planet. Or our understanding of quantum effects in semiconductor materials, in turn the basis for IC's, LED lighting, and a whole slew of other applications. Our use of smartphones & related communication technology. Or even something as "low-tech" as logistics.
Make computation cheaper, and progress that hinges on compute power, can steam ahead faster.
Another thing: as people in general get used to faster technological progress, chances are they'll be ready earlier to welcome what's coming next. When you've lived in the steam age for 50 years, electric lighting is a big thing. But when you've witnessed 10, 100, 1000, 10,000x increases in storage capacity over a few decades, a leap to 100,000x or 1000,000x is just the next step on the scale.
So the term "self-fulfilling prophecy" is very appropriate here.
> Andrew "bunnie" Huang argues that Moore's Law is slowing and will someday stop,
I think we've been hearing about the end of Moore's law for the last 15 years... inevitably, some process improvement comes along and it all keeps on going.
Yeah, it may "eventually" stop when transistors are built with just 3 atoms. Then will switch over to photonics or quantum, then some weird hyper-dimensional shit.
15 years ago they were talking about some weird 3 dimensional transistor shit.
I should use this sig to advertise my book ISBN-13 : 978-1501515132.
Yes I know technically the number of transistors on a chip is still doubling every 18 months or so; and yes that means cheaper chips that use less power. Yes that is all fine and good. But kids today don't seem to remember back when having twice as many transistors pretty much meant having twice the computing power. That 486 could do twice as much at the same clock speed as the 386 -- and the 486 was eventually going to be sold at higher clock speeds. And you didn't need to recompile anything to take advantage of all the cores they stuck in there -- even if you didn't bother to recompile anything it would still run twice as fast. Then a few years later the pentium/686/k5/"whatever they called it to avoid intel's army of lawyers" would run twice as fast as the 486 for the same clock speed and once again the chips would eventually come out with higher clock speeds.
Today you don't have to spend a lot of money on a new computer and you can still be confident that your computer will still be able to run all of the latest software many years after you buy it. In the 80's and 90's that really nice and really expensive computer (much more expensive than today's computer if you adjust for inflation) was completely hopeless in just a few years. We all knew that in some ways buying new and expensive hardware was a waste because in a few years that hardware will be so slow that it will have no purpose but to sit in the corner and gather dust. But we bought the new hardware anyway because each time we did it was like making a down payment on the future. The 80's and 90's were an amazing time to be a nerd and I just don't know if computer hardware has the same optimism as hardware of yesteryear. Or maybe it is just that I am older now than I was then.
At some point it will cease to make sense to update your computer on a regular basis. I have a 10 year old one that is fine for internet browsing and word processing. I have a friend who still uses Windows 2000 on hers - though her household does have another one. As computers get to be point of being good enough for all but the latest, most processor intense, activities, then the concept of keeping an heirloom one - especially ones designed to be upgradeable - will probably make more and more sense.
The last major, world changing thing, was the internet - some 25 years ago. Since then we've just seen it get better and better - but no real breakthroughs
Before that it was jet planes and anti-biotics - mid 50s
Before that motor cars - 1900 or so
Before that railroads - 1830 or so
Now it may be that we are waiting for the next major breakthrough.
For the past thirty years, experts have told us that Moore's Law is likely to end within ten years.
What do the experts today think? Predictions are in: Moore's Law will probably end in about ten years.
Good to see some things never change.
Irony: Agile development has too much intertia to be abandoned now.
Back in the 1960's there was a pundit/gadfly by the name of Herb Grosch, who posited a similar law about cost/speed of the various models of IBM computers. One of my first jobs at Honeywell EDP Division was analyzing the law as it applied to the 360 line. Fitted perfectly. Then we hired away a guy from IBM Hq who told us it was their pricing strategy.
One Moore's law will be a thing of the past, developers will have to take care of software performances, instead of requiring latest hardware to run badly optimized code.
For me it was reading The Adolescence of P1 in the late 1980s, with its mention of 'gigantic' 70 MB disc drives that gave me a laugh.
https://en.wikipedia.org/wiki/The_Adolescence_of_P-1
In theory, theory and practice are the same; in practice they're different. (Yogi Berra & A. Einstein)
Computers opened the flood gate to a limitless wonderland. Moore's law is simply a method of observing the radical progress in computer power. Frankly I think that five years from now we will marvel at how we got by with today's computers and electronics. I don't think we have even seen the beginning of what is surely going to occur. Maybe I'm a mindless idiot with a foggy grasp of reality but my view is that we will soon have computers capable of writing and testing programs on their own without human input at all. We will no longer design the hardware in our computers either. Computers can be used to create new computer languages that are more exploitable by machine intelligence. We already have a situation in which the game of chess is ruled by computers and even grand masters fall and fail to understand the battle plans of a chess machine. If a computer can play superb chess then how long is it before a computer can design its own software and hardware?
for flash drives ... went in to look for a 64G upgrade to my 32G-er, and they were all still way expensive (in Australia). So for the first time in a long time, look for useless files to archive and prune from the 32G-er
Manufacturers of new components need to make them "good enough" to make them generally desirable to be able to sell them. Once they are sufficiently superior to prior generations, they will sell. Of course, they also want to do so at minimum cost. Also, not too fast, to be able to sell of the old generations. As a consequence, R&D is spent to produce components that better by a sufficient factor, but no better than that. And appearing on the market in certain cycles.
As a result, you get Moore's Law.
With an hugely increased market, R&D cost matter less, and allow to identify alternatives to maintain improvements, whatever they may be. Otherwise, stuff doesn't sell. If there is money to be made, someone will find a way.
I don't know, switching from aluminium and titanium to composite materials is, such as carbon fibers is a real big deal in aviation. But this is something that you don't see and thus don't recognize. Would you know that the A350 and 787 are almost entirely made of plastic?
I agree that Moore's Law is slowing, but i doubt that we will see a slowdown in innovation. We have already seen a shift from more powerful to smaller and more energy efficient. The number of applications that need raw power are getting less and less and move into the realm of "good enough". Even in data centers you are start to see power improvement as we can do the same thing with less hardware and power consumption.
I think the next big hurdle will be network connectivity. More bandwidth, less latency.
I'll give you the smart phone; as a luddite who refuses to use one, I tend to forget their significance. Digital cameras - also true. Genome sequencing - not yet THAT significant; whilst helpful for law enforcement, we've yet to see its wider application. LCD monitors - only significant as leading towards smartphones etc. LINUX, Amazon and electric cars - nah - not that significant.
However the central experience of western life - of living in nuclear families in dispersed suburbs, travelling to work in non-agricultural occupations every day whilst children are schooled in institutions - hasn't changed qualitatively for 150 years; more and more conform to this pattern of course, but my point is that we're doing more of this - not changing those forms much.
thanks you for the info. http://www.educa.net/curso/que...
Just wanted to toss this in. Kurzweil may be an annoying know-it-all but sometimes he makes a point.
Shoot, we've been hearing about the end of Moore's Law for at least 35 years! Critics have been saying, "Moore's law only has about 5-10 years left" since at least 1980.
When will it end? I don't think we can give an accurate prediction yet. Perhaps we can have a pretty good idea when 2-d silicon will end or slow, but fortunately scientists and engineers are smart enough to keep trying alternatives.
At some point it will cease to make sense to update your computer on a regular basis. I have a 10 year old one that is fine for internet browsing and word processing
Regular yes, heirloom no. The space between physical obsoleting to the point of uselessness has and will continue to increase, but it a whole generation through which zero innovation in computers happens? less a post-apocalyptic scenario, that's not going to happen.
...A nail clipper is extremely limited in it's purpose and possible number of designs, it has a very attainable optimal design after which no substantial improvement can be made. The current and most prevalent nail clipper design is extremely elegant, it is made from only three discrete pieces each very simple in shape. It has not changed in design for about 100 years and does not need to.
Not that you'd want to inherit a nail clipper but given that it's so limited we've only just managed to optimise it to the point where there is no substantial innovation to be done to the design for 100 years... Now how complex is computer hardware and what is the scope of it's purpose, it's application, it's potential? The time to fulfil all possible innovations may as well be infinite to us, it is at least of a universal scale.
>> Yeah, it may "eventually" stop when transistors are built with just 3 atoms
Funny I was thinking along the same lines, I recall when they got to 9 or 10 atoms as being the nearest they could be, then 2 or 3 years someone came out with 8, I do like Moore's Law as a benchmark of what can be achieved. And just not in chips but in data storage and power consumption.
I really wish I could find more benchmarks on progress. it's just fun to learn stuff like this.
Oh by the way... I guessing ( using Moore's Law ) that we should have our first real space platform that is transmitting energy from space to earth, something like a laser beam or microwave beam or maybe something totally different. about 13 to 16 years from now
if you see me, smile and say hello.
And in 18 months it will be 100 years old!
>> weird hyper-dimensional shut
if you are thinking space time or something like Warp speed, not sure if their is enough power ever to achieve that in our life time ... that could happen. given I like to dream but the thought of trinary chip just seems like wishful thinking
if you are thinking LxWxH + trinary chips
if you see me, smile and say hello.
All the plastic helps with the incremental increments in fuel economy: approximately 2X better over the past 57 years. I also neglected to mention safety, which has improved a good deal more than fuel economy. That's all OK, but it's nothing like the dramatic changes that happened previous to the 707. After nearly six decades, today's planes still look very similar to a 707, are about the same size, and go the same speed.
Electronics are progressing faster then us meat puppets can deal with. We're going to have issues as electronics have the capability to take over more and more of what us humans do.
When you ask someone, what do you do? You generally get an answer of their job. it's part of our internal definition. what happens when you do nothing (and get paid nothing)?
...we all give up.
Even if we have to invest exponentially more resources into shrinking transistors, the industry is very likely to continue to invest. They will give up when the R&D costs are high enough that there is no longer any profit. But marketing has really pushed people to upgrade to new devices that they don't need, if marketing continues to do their job then we'll see Moore's Law working for quite some time to come.
“Common sense is not so common.” — Voltaire
Unfortunately high density Flash memory has a retention of months to years unless it is scrubbed. That makes it great for SSDs which are regularly used but useless for archival purposes or even as a replacement for magnetic and optical removable media in many applications.
Source? I'm interested in this.
The manufacturers do not like to advertise this so specifications are in short supply. I ran some of my own tests on various unused USB Flash drives I had laying around and none of them retained data more than a year whether powered or unpowered so I assume they do no background scrubbing. SSDs generally have better documentation and will specify something like 1 year of unpowered retention. Beware of "typical" specifications which have almost no meaning.
I've heard the term for years and thought I understood it. However, this thread seems to contain a lot of debate on exactly what Moore's Law means... I don't believe it actually has anything to do with cpu power doubling or transistor density. Can somebody clarify a precise definition?
Here is my interpretation...
If I buy a CPU today for X dollars in 18 months a CPU will exist that contains roughly twice the number of transistors that will also cost X dollars to purchase.
Yes, but where is the difference to CPUs? Many little breakthroughs in technology, most of them you don't see.