Slashdot Mirror
End of Moore's Law in 10-15 years?
javipas writes "In 1965 Gordon Moore — Intel's co-founder — predicted that the number of transistors on integrated circuits would double every two years. Moore's Law has been with us for over 40 years, but it seems that the limits of microelectronics are now not that far from us. Moore has predicted the end of his own law in 10 to 15 years, but he predicted that end before, and failed."
Moore has predicted the end of his own law in 10 to 15 years, but he predicted that end before, and failed.
So then it seems with regards to his Law, Moore has fallen prey to Murphy.
The theory of relativity doesn't work right in Arkansas.
Can we stop calling a prediction a law?
t
It will be just in time for the arrival of cold fusion.
Moore's second law: "Moore's first law will only work for 10-15 more years".
Moore's third law: "Moore's second law applies from the time it is quoted not from when it was originally uttered".
See my journal for slashdot ID's by year. Mine created in 2005. http://slashdot.org/journal/289875/slashdot-ids-by-year
There are always a few of these.
I do recall someone telling me that no CPU would ever run at more than 2GHz, as it would then start emitting microwave radiation...
Igor Presnyakov stole my hat
We will start using a new technology without transistors, but which will exhibit a similar exponential gain so long as there is money to be made from it.
In fact, now I come to think of it, ALL human endeavour exhibits exponential growth so long as there is money to be made from it. Technology is just one field where it's true. Sex (Malthus), agriculture, you name it - humans do it exponentially!
I think I'll put that on my t-shirt, and call it 'Anonymous Coward's Law'.
And if quantum computing should herald The Singularity, then it's definitely moot, since no predictions (Moore's Law included) can be made about post-Singularity computing.
Either way, when Gordon Moore eventually dies he will still be overflowing the (long)money; variable.
John
quantum computers aren't really general purpose machines and wouldn't be able to replace traditional CPUs for a lot of tasks.
thank God the internet isn't a human right.
... there's nothing fundamental about it. Instead, it's a self-fulfilling prophecy. The big players in the silicon world all use the "law" and its corollaries as their business plan. They'll likely discard a feature/product if it falls behind the curve in terms of speed. For the layperson, this "precision" may indeed create the appearance of an actual law, even though it's just an observation (similar to Malthus' "law")
The Raven
Moore's law is not about physics it's about economics. Basically the entire industry has built an economic engine that requires that growth pattern to sustain it self.
To put it another way, growth needs to be geometric not addative. that is things need to grow at x% per year, which leads to a doubling time. If the grew linearly at x += D then as x grew the proportional rate (1/x dx/dt) of x growing shrinks with time--or the doubling period gets longer and longer. Eventually it takes a lifetime before your computer is 2x more capable. Then it takes 2 lifetimes.
Why would you ever upgrade at that point? except due to wear and tear. Things become commodities and sales are based on price and other values-added. So long to intel's industry domination model.
Moore's law is also a limit too. Namely that very same growth engine will not invest twice as many research dollars to get a slightly faster doubling time. The fact that it has held steady tells you that this is so. Empirically this growth rate is the sweat spot between creating innovation at the lowest cost, and reaping a profit on it.
Indeed the only surprising thing we've seen in the consumer market that seemed (superficially) to violate this was apples' replacement of the ipod mini with the ipod nano shortly after it's introduction. They could easily have milked it for longer. But here the driver was the competition that they needed to stay ahead of.
Some drink at the fountain of knowledge. Others just gargle.
Wow, that Moore guy was so smart he outsmarted Moore.
excitingthingstodo.blogspot.com
One problem in these discussions is that different people use different definitions of "Moore's Law." Strictly the law is an observation about the increasing density of transistors (i.e. decreasing size of each transistor). However, as we all know, many people simply use the term "Moore's Law" loosely, referring to all exponential increases in computing power.
There is no doubt that we will reach a hard physical barrier beyond which we cannot shrink individual transistors any longer. This limit will be reached in a decade or two, and is probably what Moore is referring to: at our current scaling, we will hit atomic limits rather soon.
But, that doesn't mean that the exponential increases in computing power will end. There are many other things that may happen, such as figuring out ways to build microprocessors with transistors stacked in 3D (rather than having a single 2D layer of transistors), which will increase the transistor count in our computers by orders-of-magnitude. Improvements in chip designs, layouts, and algorithms are other areas that may see improvement. Specialized and dynamically re-programmable chips may also provide us with further advances. Or perhaps, as you pointed out, quantum computers will become viable and mainstream.
There is no guarantee that these more exotic technologies will work out. Yet the microelectronics industry has surprised us time and again with their ability to overcome huge technical obstacles. Thus it seems at least possible that they will deliver technology that is very much up against the physical limits of what is achievable. And with regard to those physical limits, the hard-wall the Moore is predicting in 10 years is only one aspect. There are many other ways for this technology to be advanced.
mod me funny
Next year, they'll tell us that Moore's Law will end in 5-7.5 years.
I have no idea if Moore's Law will really start to "fail" in a particular time scale (one of these times it's gotta be true, right?), but a related issue I find interesting is that CPU speeds don't seem to be being touted to computer buyers so heavily anymore. Walk into a big electronics store and look at their desktop offerings: where they used to prominently feature how many GHz they had inside (and people vaguely felt that more of these mysterious GHz was better), now the CPUs are given code names and numbers that don't reflect CPU speed: Check out this nifty X2, or the Turion 64, or ...
The new hook for consumers is the number of "cores", and once again most people have probably picked up the vague sense that having more of them inside means the computer is better. I've been told by people who might be in a position to know that it's not that they can't keep cranking up CPU speeds, but that the cost/benefit (profit-wise) stops making sense at some point because of the huge cost of implementing a new fab at a finer length scale, and we're pretty much at that point. So it makes sense that cores are the new GHz, and Moore's Law will have less and less direct impact on the end computer buyer from now on.
Maybe there's a Core Law to be formulated about how often the average number of processors per computer can be expected to double?
Moore's law will continue until THE SINGULARITY takes US ALL!!!!!!
Or at least, that's what the singularity nuts claim. Sorry people, there are limits on this planet.
Deleted
To simplify things a bit, a law is an observation, whereas a theory is an explanation. They are not the same thing, but you can have laws and theories dealing with the same subject matter.
Ben Hocking
Need a professional organizer?
I was thinking the same thing, or if quantum hasn't been quite refined as a useful science yet, at the very least, nanocomputing should be advanced enough to the point that it can take the reigns from microcomputing... That's what we like to do in this industry anyway, just change out prefixes when the technology hits a certain milestone... If you think the nano-itx motherboards are small, wait until you see the super-mega-ultra-nano-itx in 10 years, all components will be on separate cards, not quite risers, because they'll be connected via fiberoptic links directly into the bus a few microns away from the tri-tri-core CPU, the AMD Zelda-core 64000+ processor... And thanks to the miracle of nanocomputing's molecular level manipulation, food replicators will be a USB device available from Newegg... However, thanks to the miracles of the evolving software industry to meet the ever-increasing needs of users, there won't be drivers available for the new food replicators on Linux, it takes Windows 2020 over an hour to produce 1 hardboiled egg that you could have made yourself in 5-10 minutes, because it's thrashing your solid state drive (because really people, 2 terabytes of RAM is OS's recommended *minimum*), and it tastes more like a peanut butter sandwich because you don't have the latest Microsoft version of the driver for Windows 2020, because of course it makes the hardware incompatible with Windows 2020... Mac users will complain that they've had their iReplicator for 6 months already and stood in line for 12 hours to get one, and Jobs is slashing the price in half in anticipation of the release of the 2nd generation iReplicator...
he forgets his law has two variables, the number of transistors and the cost. The number of transistors might stop doubling, but there is still huge change to come with the transistors being fixed and the cost changing.
I mean, a 8 core chip would be an improvement right now, but so would a 4 core chip at half the price. Think about a world where a 80 core chip exists, and how it would change the world. It would do so again when it went from 999 dollars to 99 dollars to 9 dollars to 99 cents to 9 cents to 9 for a cent/
-You're wasting your time. Alfador only likes me.
The law speaks about number of transistors. Considering current size of a typical CPU die (about 1cmx1cmx1mm) and assuming a "reasonable" maximum size of some 10cmx10cmx10cm we have about 15 years of doubling the SIZE of the CPU (with some challenges like heat dissipation, but nothing nearly as difficult as increasing the density further) and that's not considering increasing the density any more. So even if the density reaches its limits, the CPUs may simply grow in size for a good while.
45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B2
A realistic design for a quantum computer would probably have a classical CPU that does most of the work, with a quantum co-processor. Traditional things, like running the OS and dealing with hardware I/O, would probably still be classical. The quantum co-processor would be assigned computations by the CPU that can be accomplished much faster than on the classical CPU.
This abstraction would mean that most software wouldn't have to be written with any understanding of quantum computing: libraries and compilers would be designed to use CPU calls that launch the quantum co-processor, if available.
For many operations, the quantum CPU would not be needed. But for certain tasks, it would provide orders-of-magnitude speed boosts. If quantum co-processors became commonplace, we would see improvements in all kinds of parallel-processing tasks (matrix operations, simulations, graphics, maybe even search?).
I predict the number of predictions of the end of Moore's Law will double every six months.
We've secretly replaced Slashdot with new Folgers Crystals - let's see if it notices.
Nobody will ever need more than twice the computing power he already has available on the moment he makes this assessment.
Linux user since early January 1992.
Luckily, there are enough geeky pedants on slashdot to make up for the fact that the editors have actually messed up this totemic bit of geek lore.
Moore was/is a technology manager, and his law is a management law. It says the number of transistors that can be economically placed on an integrated circuit, i.e. the transistor density of the price/performance "sweet spot", will increase exponentially, doubling roughly every two years.
The original refers to "complexity for minimum component cost", which emphasizes the economic aspect of it even more strongly.
Moore's law has never been about what's possible, it's always been about what's cheap.
2*3*3*3*3*11*251
If you accept the statement I just made about moore's law being sustained because of economics then here's a corollary which makes an observable prediction.
Moores law stays fixed because the industry invests enough research dollars--and not one dollar more-- to keep it at that rate. Their entire economic model is built on this.
Therefore, if we every do reach a point where we simply are running out of available physics and computer science (multiprocessing) then the first sign of this will be an increasing fraction of research dollars spent to sustain moores law.
Plot the industry's margin, smooth the curve, and you will be able to extrapolate to the point where the research dollars cross the profit line. somewhere shortly before that is when moore's law will end.
The only way that would not be true is if the nature of innovation changes from frequent small leaps to massive leaps spaced far apart.
Some drink at the fountain of knowledge. Others just gargle.
Homer: Lisa, in this house, we obey the laws of thermodynamics!!!
All ideas^H^H^H^H^Hprocesses in this post are Patent Pending. (as well as the process of patenting all postings)
Secondly it's not so much a "law", as a consequence of how long it takes to amortize the cost of a fab plant.
Thirdly, it's tied to 2-D circuit layouts. If and when 3-D IC technology becomes practical, then all we need is 2^1/3 percent or about 22% linear shrink every year, which is somewhat more maintainable for a few more generations.
That's 32 times as many transistors... whereas today you can get 4 cores on a CPU in under 300mm^2, you'll be getting 128 cores in 2017 (simplistic, you'll get a variety of generic cores, and application specific cores, and per-core improvements will increase their size, so say 32 generic cores and 32 application specific cores).
If it's 16 years, thats 256 times as many transistors. 256 generic cores and 256 application specific cores in 2024? Let's not even imagine the per-core speeds! It's all pretty exciting, and I'm being conservative with the figures here.
Of course, applications will grow to utilise this stuff, but more and more tasks are getting to the point of 'fast enough', even despite the bloating efforts of their creators. Even if there is a 10 year hiatus in process improvements after 2024, it'll take some time for the applications to catch up apart from certain uses. If those uses are common enough, there will be hardware available for it instead. Of course if only Intel and IBM have fabs that can make these products, because the fabs cost $20b each...
It's not a law, it's simply an observation that within Intel, that's more or less the rate of progress. As we saw with the P-4 chip the problem we bumped into was not Moore's Law, but the laws of thermodynamics. So we found a good enough reason to go to multicore CPU's. Eventually though you do bump into Albert Einstein. In 1 billionth of a second, light travels about 1 foot so the entire circuit length from end to end, in order to have a switching frequency of 1 billionth of a second, has to be less than one foot.
Why buy a computer this year, when I can get a faster one next year?
The fuzzy logic behind not buying a computer due to Moore's Law.
Somewhere, once a upon a time, I saw an article that took the opposite approach: They worked out what the absolute maximum transistor density was, and worked out from that when Moore's Law had to end. They figured one transisitor per Plank-unit, in a spherical computer. (Where the clock speed is proportional to the size of the sphere, governed by the speed of light.)
IIRC, it ended up something like 150 years in the future.
'Sensible' is a curse word.
Exactly. Whenever one process technology reaches its physical limits, we get a new one, because the new process makes money. X-ray lithography, chip stacking, 3D circuits, and eventually nanotech will all keep us on the Moore's law path probably for the rest of my life, at least.
Ye be forgettin' one thing, matey, they be makin' multiple cores now. Eventually we be lookin at distributed computing on an individual platform. Ye may be layin' claim to Moore's law applyin', but it be tenuous a claim at best. The paradigm be shiftin' away from the domain of Moore.
A feeling of having made the same mistake before: Deja Foobar
That's nonsense. The industry grew around the physics, not vice versa. The fact that the industry is predicated on a constant improvement of speed and complexity is because such a thing is achievable in microelectronics, certainly not because microelectronics is the only industry where such a thing is desirable.
I mean, who wouldn't want cars to become twice as gas efficient (without losing power) every 18 months, ad infinitum? If such a thing were technically possible, it would happen, because all the car makers would jump on the gas-mileage bandwagon to get ahead of their competitors.
Who wouldn't want the amount of food that can be grown per man-hour to double every 18 months, so the price per pound of beans and broccoli fell as fast as the price per CPU cycle of computers? If such a thing were possible, it would happen, as every farmer raced to lower his costs of production and undersell his neighbors like crazy, earning millions.
In very few industries other than microelectronics has anything like Moore's Law applied, and that's not from a lack of economic incentive, but from the plain uncooperativity of Mother Nature. You're arguing backwards, from effect (the economic structure of the industry) to cause (the physical nature of microelectronics).
They only change in ways that are generally not possible to anticipate, hence which haven't been predicted.
And of course they would. Technology, like the stock market or the weather, is inherently a chaotic system over a certain characteristic timespan (1-2 weeks for the stock market and the weather, 25-50 years for technology). That is, over the characteristic timespan very small causes can produce enormous, system-wide effects, what you might call the butterfly wing flapping causing the hurricane phenomenon.
For example, a couple of guys (Jobs and Wozniak) screw around in the garage in the early 80s, trying to put together a really cheap personal computer. That's a very small cause. And twenty-five years later, it has had a giant effect: iMacs and iPods and iTunes oh my. Problem is, there was no practical way in the 1980s to distinguish the small cause that mattered (Jobs and Wozniak) from the other 50 zillion small causes that didn't matter (the other 50 zillion pairs of scruffy entrepreneurs in garages whose brilliant idea went nowhere).
This is why predictions of the future out more than 50 years usually end up looking hilarious in hindsight. When sf writers of the 50s and 60s predicted the present, they projected the dominant themes of their time (spaceflight, atomic physics, the struggle with Soviet Communism). They did not -- and could not -- realize that all three themes would pretty much abruptly and surprisingly come to an end in the 90s. When present writers predict the future, they project the dominant themes of our times (e.g. networked computing). It's very likely these projections, too, will end up wildly wrong. Networked computing is likely to become as humdrum and static as telephony within the next half-century or so.
I'm not sure if this is the same article that you saw previously, but this paper discusses that topic:
Seth Lloyd, "Ultimate physical limits to computation" Nature 406, 1047-1054 (31 August 2000) | doi: 10.1038/35023282 (for those without access to Nature articles, this arXiv preprint appears to be the same article).
The article reviews the absolute maximum limits for computation, based on current understanding of thermodynamics, relativity, and quantum mechanics.
The basic conclusion of the paper is that a theoretical 1 kg computer (confined to a volume of 1 liter), operating perfectly at the edge of what is physically possible could compute 10^51 operations/second on 10^31 bits of information (as compared to our current computers: 10^10 operations/second on 10^10 bits). Naively scaling Moore's law from current sizes, this suggests that we will reach such limits in 250 years. Of course the paper repeatedly points out that this is for an unrealistically 'perfect' computer, that is somehow able to perfectly organize all its internal matter solely for performing the computation at hand. For instance when running a computation it effectively has a temperature of ~10^9 Kelvin, which is considerably hotter than any known material could withstand.
Nevertheless, it's interesting to see what the fundamental principles of relativity and quantum mechanics indicate as a boundary for any sort of computation. The article is an interesting read.
w00t
Moore's 2nd law is that Moore's 1st law is going to come to an end in about 10 years. Always.
Do you have ESP?
Moore is being short-sighted about his own law. It's not about silicon. If you extraploate backwards from the first integrated chip you see that "Moore's Law" has been in effect for over 100 years. It started with manual switches, then moved to electric motor switching, then to vacuum tubes, then to transistors, then to integrated circuits. Every one of those mediums has been subject to and demonstrates Moore's Law. Graph it and you'll see. It's a perfect logarithmic line. Every time the method itself peaks of its own accord a new medium is found which can continue the progress. (Any familiar with the growth of telco equipment can see this in the switching systems: Electric switches to step systems to crossbar to ESS.) If IC does run out, there is a future of possibilities: holographic, quantum, bio, etc. Moore's Law is like the Energizer Bunny. It just keeps going.
How about a moderation of -1 pedantic.
Moore's law says nothing at all about the speed of a processor, or of a program. It only says the number of transistors will double every 2 years. The fact that performance benefits have traditionally been had by adding transistors does not mean this will hold. In fact today the performance of most applications is no faster on current computers than top of the line computers from 2 years ago (it's definitely not twice).
Phil
Moore's Law describes a CPU speedup that died at least 3 years ago (all other legalisms aside).
To wit: I bought a laptop in 2003 with a 2.2 GHz 32 bit P4. According to The Law, by 2005 CPUs on comparable laptops should have run at 4.4 GHz, and by today they should zip along at 8.8 GHz. But in fact, no commodity CPU runs at that speed nor even *half* that speed.
And don't you believe the claim that multicore or power throttling compensates for or explains The Law's "failure to thrive". The fact is, the industry is no longer delivering CPUs whose SPECMarks/FLOPs/etc (AKA performance) is rising at the rate that they have for the previous 20 years. I tell you, "Moore's Law is pushin' up daisies. It's a DEAD parrot."
What's puzzling to me is that while this emperor is clearly naked, for some reason, nobody wants to admit it. Why not? Are we afraid that sexy soothsayers like Ray Kurzweil or Rod Brooks will be regarded laughably when they forsee cool stuff like The Singularity or robots possessing human-level cognition, brought to us by the inexorable exponential march of Moore's Law? Or do we simply dread the day when we have to depend entirely on advances in *software* to deliver our next high-tech fix? Perish forbid *that* thought.
Well, we'd better get used to it, the emperor is naked *and* dead. There's a new emperor in town, and Moore's Law 2.0 depicts a future that looks a hell of a lot like the past.
Randy
IC's today are made photographically, on a flat surface. Manufacturers keep working to reduce the area needed for a component, be it transistor, resistor, capacitor or trace wire. We already know from lab work what the minimum possible sizes are for each basic component. We've come up on the minimum possible size several times in the past. Each time, it was related to the possibilities of the light source we were using. Now, we are up there in the extreme UV range, and have minimum feature sizes that are actually smaller than the wavelength used. The best commercial plants use a 45 nM wavelength. At about 30 nM, the traces (on chip wires) become unstable, and may no longer be conductors. That is a fundamental limit that clever plant engineering will not be able to surmount. Current commercial plants are using a 60 to 90 nM min. feature size, if memory serves. That means we have about 6 or 7 doublings (each doubling is about a 70% reduction in feature size and takes 2 to 3 years t realize.) That gives us 12 to 20 years.
Going to still smaller wavelengths means that the photons pack more punch. It's like trying to play billiards by shooting the cue ball with a high powered rifle. You get pieces of cue ball everywhere. When random photon collisions are pushing random atoms by several dozen radii, your nice ordered atomic lattice becomes a horrid mess. we are nearing the limits of what nature allows for photo lithography now.
Increasing chip size is not a viable solution, as the full wafer is used now. Increase chip size, and yield drops quickly. Yes, they could double the size of the chip to increase transistor count, but that would mean increasing the cost of the chip by 4X. That's not he direction we want chip cost to go.
Off in the distance, there are more real hard boundaries, beyond which no amount of effort will yield additional benefits. One of those is component size. Minimum transistor size is 7 atoms (it's been done). Minimum diode size is about 5. Minimum trace size varies with material. The best I've seen is benzene, at about 6 atoms width. Keep in mind that at room temperature, benzene is a gas. It's going to be very hard to make wires of the stuff. We really need a solid. Aluminum, silver, gold, all have been used, and all need to be 30 to 60 atoms wide or more, and several thick to be even a poor conductor. Some creative metallo-insulator engineered materials might allow for smaller trace sizes, but probably not. Please note that this is still smaller than buckytubes, which are also as tall as they are wide, creating other connection problems, so don't peddle that as a panacea. That means that the trace sizes required will probably be the final limit. Real capacitors are larger than the traces, but their size is really controlled by the number of electrons needed to operate the transistor/switch. I'm still betting on the traces as establishing the limit.
Heat dissipation is also a problem. It gets to be more of a problem as densities go up. Current best designs are operating half way to melt now. switching to silicon carbide would let us go hotter, say 400 to 800 C. Diamond/graphite bases would let it get higher still, though diamond heated to 1,200 in an oxygen atmosphere isn't going to last very long. Need some creative packaging there. Heat dissipation is the real reason we can't go 3D. The systems that tried to be true 3D, or near to it, all relied on the chips being immersed in some coolant and having channels for the coolant through the chip. Liquid nitrogen cooled some that IBM did a few years ago. bubbles were a problem. move the coolant fast enough to transport the heat before bubbling and erosion is a problem.
Some of these issues can be fixed, some can never be fixed. So, when we are fully 30 nM size with our components, it all stops. It's a problem with the wiring. Solve that, and we would be close to being able to compute with atoms. But, with what we think we can do now, the shrinkage stops in about 20 years.
Enjoy it while you can.
Looks like you
Everybody knows 3 people with my name.
I agree. The real point is that Moore's Law is not dependent on Moore, nor on silicon. If in the past researchers had fixated on the vacuum tube, they never would have reached beyond the vacuum tube paradigm to make the advances that happened. I am encouraged by the results other research labs have already achieved with these new mediums. It's not so much that they still need to be invented as much as it is that their discoveries need to be developed. I think it was William Gibson who said, "The future is already here. It's just unevenly distributed."
How about a moderation of -1 pedantic.