Its kind of good for both Physics and innovation. Moore's law has stalled a lot of promising technologies that should have received funding but ecomonics stopped all such technologies. Hopefully some other material than silicon will see the light.
You could say that people have already pulled their head out of "sand" ( i,e. silicon ) and looking at other materials with higher electron mobility and lower leakage. Did the ex Intel guy at ARPA ( whose name I did n't recognize ) get into any of those novel materials ? If and when they move from Lab to Fab, they might even upset the current pecking order among SemiCos ( both IDM and Foundries ).
Good Physics have a way of disrupting extrapolations - both for and against Moore's Law.
@ Garcia - Lasheras : selective presentation of data ( graph showing saturation of clock rate ) only damages one's own credibility. As it should have been clear to you this is a Technical Forum and participants are well aware of clock - rate vs multi - processor arguments as options to deliver a certain GFlops. That train has long left the station.
@chipmonk0: "selective presentation of data ( graph showing saturation of clock rate ) only damages one's own credibility".
Thak you very much for your value judgement...
"As it should have been clear to you this is a Technical Forum and participants are well aware of clock - rate vs multi - processor arguments as options to deliver a certain GFlops"
I'm pretty aware of this, but not everybody share the same skills or knowledge. The graph is included in a blog that I wrote for the All Programmable Planet community, in which I tried to give a very simple physical explanation about why clock rates doesn't fit with Moore's law beyond a process limit despite the fact that transistor number does. If you are really interested, here you have the full history.
The "selective presentation of data" I expose was obtained & kindly shared by Dr. Colin Gillespie, from the Univerity of Newcastle -- UK. In his original blog, he explains how he collected and analyzed the raw data -- you can check it in deeper way if you want to.
"That train has long left the station"
I started working in the clock related problems in CMOS processes in year 2001.
From your comments, I deduce you are pretty interested in the Moore's Law demise or survival topic. I plan to be in the related EETimes week in review chat tomorrow... maybe shall I see you there?
Was n't it you yourself who reported a couple of weeks back on Samsung's 3D NAND -- stacked one cell on top of another on a single wafer ?
With time, such 3D technology will spread into processors and memory - processor combos as well. Though the cost per transistor may increase in such stacked devices due to cost of addition of isolation & vertical interconnects etc., the cost per system & power consumption would go down radically due to electrical reasons. How is that for not having to wait for adequate throughput from EUV but still stay on the growth path ?
Balanced, comprehensive and well-reasoned coverage is what we will keep depending on you for.
@rick merritt >> Regarding graphene (along with many other nano-particles) and toxicity at the human cell-level has been raised in the past few months (please google it and you may find the topic interesting). I also have issues with the "fact" that the only byproduct of hydrogen fuel cells is supposed to be water. Donald Rumsfeld 'poetry' regarding the Knowns and the Unknowns may hold true here (please google it and you may find that interesting also). Cheers
Indeed the economic stop seems to be a far more predicatable one than the technological stop. Not only will very few players be able to afford to build a fab or even tape out an IC in 7 nm or 5 nm or whatever the last CMOS node is, but it is also likely that very few ICs will command the enormous volumes required to justify the expense of developing them and getting them released to production.
The way it is going now as regards to the $$ required to build the next generation fab, it may very well take $10B+ to build a 5nm fab, affordable to only a handful of them, perhaps three? Any guesses? TSMC...
There seems still some way to go in terms of nm transistor size, and already we are expanding into 3 dimensions. Have the esperts taken this into account? My feeling is the 3D expansion will carry us a fair bit further. Could someone who knows more than I do (which is not much) comment?
The world is full of experts and speculation is cheap. We could fill a magazine with reports that "Moore's Law Dead by Year X, Expert Says." I've been reading such reports for years - and many of the stipulated dates have passed. It seems to me that we're playing a game of chicken. I guess Moore's Law will be dead by 2113, you guess 2063, someone else guesses 2022 (which is still a long way off) and wins the headline lottery for this week. It isn't news. The one positive outcome of this speculation is that it re-energizes the inventors to find one more creative way to eek out another doubling of performance, which benefits us all.
Babson's Daniel Isenberg authors "Worthless, Impossible and Stupid: How Contrarian Entrepreneurs Create and Capture Extraordinary Value" Since it was published by Harvard Business Review Publishing, it has risen to the top ranks on Amazon books and has been positively reviewed or featured in the Economist, USA Today, Wall Street Journal, Financial Times, Forbes TV, Bloomberg, USA Today, Ventureburn, Entrepreneur.com and a number of radio programs.
@pseudoid: Hey if there's a contrarian out there with a new device manufacturing technology that promises exponential growth for the next 30 years, let that person stand up and speak! Until then, umbrella anyone?
@DRQuoine: Indeed as Intel's mark Bohr used to say "People have been predicting the end of Moore's Law in ten years for 20 years" The difference is now they are saying in 7-9 years, and I suspect that time frame will only shorten.
I disagree with Colwell"s statement,"when Moore's law stops it will be economics that stops it, not physics". Based on the transistor physics, the short channel effect or transistor leakage current will increase as the transistor is scaled. Today we have only three process technologies: Bulk Si, FDSOI and FinFET. Bulk Si technology is in high volume manufacturing at the 28nm node for several years by major semiconductor companies, and may extend to the 22nm, but with excessive transistor leakage current, thus definitely not beyond the 22nm. FDSOI is invented by IBM more than 10 years ago, but not manufacturable at any technology node yet mainly because Soitech, the largest SOI wafer supplier can't deliver 7nm thin SOI that is required for the 28nm node. It means that FDSOI is already dead and Bulk Si will be dead at the 28nm, not because of "the money" for manufacturing but because of device physics. FinFET is the only technology in volume manufacturing today at the 22nm by Intel, and the 14nm will be manufactured sometime in 2014. The beauty of FinFET technology is that it can extend to the end of scaling or even to the 1nm node according to FinFET physics. In order to overcome the short channel effect the Fin width, W is equal to the gate length, Lg is required. Or W(Fin width)=Lg(gate length). It means that for the gate length, Lg=3nm, the Fin width W=3nm is only require to overcome the short channel effect. The Fin width, W here is equivalent to the channel thickness for the conventional transistor. Therefore, FinFET will be able to extend to the 3nm node by 2030s beyond the 7nm by 2022. Multiple exposures are used today by Intel for the 22nm FinFET manufacturing. With availability of EUV and 450mm wafers possibly at 14nm, 8nm and 3nm nodes the manufacturing cost per die and per transistor will be significantly reduced. There is no alternative to FinFET today. Moore's Law will be alive to the end of FinFET scaling. SKim
I believe that as fewer resources are directed at improving vanilla CMOS due to the slowing of Moore's law, other related technologies will flourish. It's beginning to happen already. The author is correct when he says to follow the money. The money has started to flow away from CMOS to MEMS, photonics, flexible electronics, and a plethora of other technologies that are poised for fast growth. If you were a wealthy venture capitalist today where would you invest your money?
Moore's observation and move on with our lives. The death of Moore's law has been forecast repeatedly for the last 20 or more years. So what if it eventually happens. We will have EXTREMELY dense silicon (or whatever you want to call it) by the time it happens (to match the EXTREMELY DENSE pundits who worry about it). Move along folks, nothing to see here ;-)
Few years back, there was an analysis of inventions/innovations/discoveries over the past centuries; dating back to the Mongolian era. I wish I could find that article which had discussed the revelations about what it takes to enable the paradigm shifts that have occurred. But one of the findings of that analysis was the shocking [to me] fact that they actually happen in spurts and in bunches during a small period of time. The analysis had further gone and found out that these cycles were a confluence of events that were not just limited to drastic social changes and congregation of equal minds being collocated in a small area of the world, just prior to such discoveries. Google searches will lead to many different hits on such happenstances. I hope I have provided the fish hook for those interested in fishing further (or are hungry enough)!
It may be that we are in a period of time that is the lull before one of those storms are about to happen again.
Thanks for your article. You know, this is a joke. Why don't these so call experts spend time developing a new concept that can help society instead of focusing on what has been around for years. This semester I am teaching a college level Business Comptuer Application course and next week the lecture topic will be, you guessed it, Moore's Law. Moore's Law is not dead and Moore's Law will not be dead as long as we continue to teach it within the educational system. Moore's Law, Moore's Law and more Moore's Law.
Charles, what you said is a philosophical reality while in practice we all know the Moore's law as we know it today will hit the wall both economically and technologically. Though we might discover some way to satisfy our ego and keep improving the hardware performance.
I think other articles have morphed Moore's law into more of a functional "double the transistors" rather than size. This allows 3d developments to fit into the concept and is much more usable. I'd think an "expert" would realize this. Obviously, you eventually get down to atomic sizes and it becomes difficult to get there from here. But transistors per unit area is would account for 3d stuff.
Well, if philosophical reality is not true, I think we all professor's including those at Harvard, Yale, MIT and other colleges and universities around the world need to stop training future engineers. Will that happen in 20 years, no. A law is dead when there is a major movement to change it. There's no movement to change Moore's Law. As a matter of fact, this is the first time I have heard of such statement. Now, I am willing to add to or enhance Moore's Law to meet current times, but to say Moore's Law is dead... please. Sometime in life you have to face reality. Moore's Law is current and we have no intention of changing it. Moore's Law helped lay down the foundation. You don't remove educational foundations out of the curriculum.
I think one thing people are forgetting is that Moore's Law is not a law. It is not a law in the sense of the law of Gravitation for example. It is just a statement he made in an article many years ago that people latched on to. The laws of physics don't change, but Moore's Law (that is not a law in the scientific sense) can and will change.
Philosophy applies to ethics and politics. Not to science and technology.
Let science and economics decide whether moore's law is dead, alive (or both).
(When I have a leak in my bathroom, I don't call an electrician, I call a plumber.)
Right now, physics tells me it's impossible to create a transistor with a channel length < an atom's width, and economics tells me it's impossible to invest more than the world's PIB in fabs.
If we want to be able to produce the next generation's smartphone or computer and not having it take more space than we have available in any limited place (a flat, a warehouse, a country, even a planet: take your pick), then at some point, we know Moore's law is going to come to an end either because the device is going to be too big to be practical, or because we don't have the money to build it)
I agree. The fact that relativistic physics is needed to explain the observed orbit of Mercury does not prevent us from teaching Newton's laws of motion. A philosophical question is whether or not we SHOULD teach Newton's laws. Note that we still refer to those equations as Newton's "laws", though they are known to inadequately describe observed reality. In other words, Newton's laws of motion describe motions that are physically impossible. It is quite possible that Einstein's laws will also be shown to have inaccuracies. So, another philosophical question is just how close to observed reality does an equation have to be in order to be considered a law?
Moore's Law is dead predictions began about 10 years ago. Sooner or later, somone will be correct. Personally I think there will be another technology to replace CMOS, but that doesn't mean Moore's Law will continue to be valid.
this discussion / debate on Moores Law is getting ridiculous like medieval monks arguing about how many angels can dance on the head of a pin. i say ENUFF. The reporter ( @ Rick ) should make an honest effort to stop this wastage of time by posting the original slides of this Colwell fellow at DARPA. He is after all a Govt. employee and we pay his salary, he owes the semiconductor community the technical details behind his proclamations. let us see for ourselves how technically sound his presentation was.
Blog Doing Math in FPGAs Tom Burke 4 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...