Moore's Law has ten years to run, predicts physicist
Renowned theoretical physicist Michio Kaku has predicted Moore's Law will run out of steam within the next ten years as silicon designs run up against the laws of physics. "In about ten years or so we will see the collapse of Moore's Law," he says. "In fact we already see a slowing down of Moore's Law. Computing power cannot …
This topic is closed for new posts.
Re: And not a moment too soon!
Software is getting slower faster than hardware is getting faster.
But this is clearly nonsense. If it were true then the latest games consoles would be incapable of playing Manic Miner.
That is worrying.
Because ... ?
@ Some Beggar
What Wirth means is that for a given task, the current software needs vastly more resources (CPU and memory alike) than similar software years ago.
Why is this worrying? Because it suggests that we could get by on much leaner compute capacity for many mundane tasks. It means machines that still work fine have to be replaced when the software is updated, and the minimum specs are upped again. This is ultimately wasteful. It also means that bigger server parks are needed for a given workload. If you can make code more efficient, less hardware is needed, and less energy is wasted. Mobile computing (like embedded) can give an impetus to leaner programs, simply because cutting clock cycles can cut battery usage.
Re: @ Some Beggar
What Wirth actually said was this:
"Do increased performance and functionality keep pace with the increased demand for resources? Mostly the answer is no. "
Which is simply a statement that we become more profligate when we have more resources. There are variants of this truism dating all the way back to the Hebrew Bible. It's why Americans build gigantic cars with inefficient engines - they have big roads and cheap petroleum. The only time this becomes a problem is when the resources start to be throttled. This isn't going to happen any time soon with mips and memory. Wirth wrote this nearly 20 years ago and the computing industry hasn't collapsed as far as I've noticed.
The paraphrased version that you quoted is simply nonsense.
re: Michael H F Wilkinson
I'd like to agree with you about mobile computing but it's not what I see when I look at smartphones. For every improvement in battery tech and efficiency gained through die shrinkage I see an increase in clockspeed to deal with ever more bloated software.
Re: re: Michael H F Wilkinson
And these smartphones have processing capability (in terms of what the user can actually do - not simply in terms of what the silicon can theoretically do) that is well beyond what a desktop PC would have had back when Wirth made his oft-misquoted statement.
So you can now do things on a battery-powered pocket device that were impossible on a mains powered box that lived under your desk twenty years ago. How exactly does this demonstrate that software has bloated faster than hardware has speeded up?
Clark's Law trumps Kaku's prediction, thus extending Moore's Law
Arthur C. Clarke's laws of prediction:
1) When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.
2) The only way of discovering the limits of the possible is to venture a little way past them into the impossible.
3) Any sufficiently advanced technology is indistinguishable from magic.
Michio Kaku is certainly distinguished. He's also recently turned 65, and that makes him elderly. See Rule 1. Moore's Law will shift gears, but it will continue further than expected. Just like last time. And the time before that.
Re: Clark's Law trumps Kaku's prediction, thus extending Moore's Law
"Michio Kaku is certainly distinguished. He's also recently turned 65 and that makes him elderly."
Steady on with the elderly bit! - I'm 66 and I consider myself only a teeny bit old (though my children and grandchildren might hold a different opinion!
Moore's Misnomer
It actually refers to integrated circuits, not silicon, and it isn't a law, but who cares. As mentioned above, there will inevitably be discoveries that ever push the horizon a few years forward, until the day we realize we've had enough math.
Re: Moore's Misnomer
Strictly speaking the Second Law of Thermodynamics isn't a law either, but it still seems to hold.
And don't talk about enough math.
The path beyond the next few Moore's Law process generations has always been quite murky.
However, just because previous predictions of its demise have been wrong doesn't mean current predictions are also wrong.
If I understand correctly, a *lot* of the innovation in the past decades has been about pushing the need for big changes (like EUV) forward into the next process, then the next, and so on. How long can that go on? Even if the technologies have passed far beyond my understanding into the realm of apparent "magic", Moore's Law is not a magic law operating outside the physical world - exponential growth will always slow down at some point.
There's Mores Law Part 2 though.
Part one states that every 2 years or so, the transistor count in an IC will double.
Part two state that every 2 years or so, someone will incorrectly predict the end of part one.
They are already working on Silicon replacements
http://www.research.ibm.com/nanoscience/nanotubes.html
Re: They are already working on Silicon replacements
Yes, the materials and production processes will probably change. The economics of chip development have changed significantly in the last ten years due to the cost of making the machines that make the chips. There are fewer suppliers of the lithography machines than there were ten years ago which is driving up the cost of each new generation. At the same time even Intel's margins are starting to come under pressure as it's designer hardware struggles to differentiate itself from the commodity ARM clones.
The number of companies entering the nano-technology and additive manufacturing to address some of the same problems is increasing following the. If they can get it right printed OLED screens may be the first children of this revolution.
Worth noting that IBM and Samsung may well be right in not having an asset-light strategy in this area.
Am I the only one?
I don't have that much respect for Michio Kaku. Yes, he's got a good brain and yes he's good at dreaming up fantastical stuff. But he has a penchant for seeking media attention and spouting crap about area's of expertise he's not nearly qualified to talk about. (Like around the time of the whole Fukushima debacle)
Maybe I've missed it, but I haven't seen anything useful done by mister Kaku. He keeps "thinking up" wonderful new crap that's either highly improbable or just plain wrong and impossible from an engineering standpoint. (using materials and methods he suggests atleast. And often those things aren't even all that original)
I wouldn't be bothered if he just kept making dull TV shows, but he keeps going to the media with these kinds of statements, that, to me, just seem like he's seeking media attention again.
Re: Am I the only one?
I couldn't agree more. If you've ever seen Koo-Koo Kaku speak in person, you'd have been witness to a spotlight-seeker who'll say nearly anything to grab the attention of folks with weaker scientific chops than he has. Sure, he's educated and articulate, but his pronouncements are increasingly tales told by an idiot, full of sound and fury, signifying nothing.
Also, remember that he's a physicist and/or astrophysicist, not a materials science guy. If you want to know the future of microprocessor technology, talk to microprocessor technologists. I think the Reg has published some stories about this matter in which they talk to people who actually know what they're talking about.
Hang it up, Michio – your 15 minutes of fame are over.
Moore? Really
Why does Moore continue to get credit with the physicist's prediction of processor speed doubling annually? It was not Moore who discovered or predicted it but who made the prediction well known by non-physicists!
They've been saying we're about to run into physical limits for as long as I can remember. It's a bit like fusion... always just around the corner.
Doubling CPU cores is also doubling transistors
In the end the article says that manufacturers look for for other ways of expanding CPU power, such as increasing the number of cores in a CPU, etc. The CPU cores have been doubling every 2 years over the last 5 years or so - how is that not doubling the number of transistors? This physicist obviously isn't an IT professional who would otherwise know that the next step of scalability is not vertical, but horizontal.
Re: Doubling CPU cores is also doubling transistors
One of the problems that chip designers have is how to use the vast number of transistors that can be fitted onto the large die-sizes at the smallest scale.
They got to the point where more registers, more cache and more instructions units in a single core was not making for faster processors, so they then started using the still increasing transistor budget to put multiple cores on a single die.
There is a lot to be said for a large number of cores on a single die, but this has it's own problems with access to memory, cache coherency between cores and I/O.
Another avenue is putting disparate processors (like GPUs) on the same die, or even System on a Chip (SoC), where all of the functional elements (I/O, Graphics, memory etc) of a complete system appear on a single piece of silicon (think what is going into 'phones and tablets).
In my view, to make use of the vast scale of integration, it's about time we had a fundamental rethink about how processors work. I don't have any new ideas, but I think that listening to some of the people with outlandish ideas might be worthwhile in coming up with a completely new direction to investigate.
Graphene will be key to take Moore's Law strain
As Gronkle says, Graphene is already spinning up & ready to take over from where Silicon's physical limits are reached. With the atoms being so much smaller than Si and some, I believe, wondrous other characteristics at such a small scale, I think we have a way to go yet, maybe even until quantum chips do come online in quantity.
am i the only person to know that in 2011, MIT lab researchers already produced non-silicon chips that can run at 100GHz or more? sometimes i feel i am. we already have this technology and 99.99% of the world has no clue it exists.
it's just the manufacturers don't want to put in the R&D and production costs when they're making enough $ with silicon.
I'm in this 99.99%
would you please enlighten me what this technology is and, perhaps, why isn't it commercially used?
Sadly, Wirth's Law will keep on going
"Software gets slower faster than hardware gets faster"
Luckily for us, the annual increments in processing power have masked the continual degradation in software performance. If Moore's law does hit the buffers, someone somewhere is going to have a 50 year project on their hands to fix all the inefficiencies, crocks, bad design and unoptimised code.
Maybe once there is an actual incentive to write fast, efficient code instead of the traditional "do whatever it takes, just deliver it by Monday" approach to software development, we'll have a renaissance in programming. If we're lucky, we may even find that once people start writing well-designed, reliable and efficient software that programmers start to win back some professional esteem and respect, too.
Re: Sadly, Wirth's Law will keep on going
Software is ALWAYS about doing what it takes to run on the target hardware. It's all well and good being nostalgic but people used to hand-optimise ASM because they had to to make it run fast enough, not due to some sense of craftsmanship.
Those old-school skills are fun to learn and good to know, but you learned them because it was necessary to get the job done.
Re: Sadly, Wirth's Law will keep on going
It seems to have escaped your notice how much work has been done on compilers in the last few years which do an increasingly good job of removing many of the inefficiencies. That and the ability to shift tasks to hardware implementations (encryption, signal processing, video compression and decompression).
Re: Sadly, Wirth's Law will keep on going
"Maybe once there is an actual incentive to write fast, efficient code instead of the traditional "do whatever it takes, just deliver it by Monday" approach to software development, we'll have a renaissance in programming. If we're lucky, we may even find that once people start writing well-designed, reliable and efficient software that programmers start to win back some professional esteem and respect, too."
Whining cry baby.
how much work has been done on compilers in the last few years
People have also been saying that for years. It's undoubtedly true but a good optimiser can still beat the hell out of the compiler through knowing the context and stuff like SSE. The claim "modern compilers can't be beaten by hand coding" is simply bogus - however the truth is it doesn't normally need optimising.
While I wont go so far as to predict the time scale I will say this, Moore law will end one day. There is a physical limit to the information you process with a given quantity of matter.
Re: You're assuming
In order to be useful to us material humans, at some point the processing has to interact with the material universe.
Or are you suggesting we have to transcend to some higher plane of existence to get around Moore's law?
Chinny Reckon?
"the first proper quantum systems won't come online until late in the 21st century"
I'd give it 20 years meself, and I'm not known for my optimism.
Re: Chinny Reckon?
Much uncertainty in quantum computing there is young padowan
Re: Chinny Reckon?
> "the first proper quantum systems won't come online until late in the 21st century"
> I'd give it 20 years meself, and I'm not known for my optimism.
And even then they'll be online and not online at the same time (...er, if I've understood it correctly?)
We havent even started yet....
when I first started making micro chips 32 years ago there were lots and lots of techniques being looked at that haven't been re-looked at that could easily keep Moore's law going for another decade or two.
I have often wondered if they were really dead ends, just swept aside by the usual monotonic rush, or possibly as they were not patentable any more and so wouldn’t give any more than a moments advantage to anyone who developed them to fruition.
But as has been alluded to earlier in this thread Moore's law has an 18 month doubling and Wirth's law has software doubling bloat in 17 months.
Me - I'd be quite happy with a 4 core arm machine with a socket for adding more when I need to.
It wont run microsoft office - so what? My car wont pull a plough.
Naaa, there is still new technologies coming up
stack of graphene and silicene techno-wizardry combine with nanotubes coduits and "on the chip" dynamic (fast moving fluids) heat transfer solution is my bet for the future.
we still are at infancy on term of internal heat transfer solution that is at the end of the day the main physical limitation.
Software bloat
I am personally going the other way, trying to squeeze a little more out of it.
Word processors need little hardware, if they do they are poorly written.
What happened to diamond chips?
I'm sure I read about those years ago. As I recall they were very heat tolerant, so we could go back to the GHz race (well for desktops, heat might continue to be an issue for stuff you have to be able to hold).
Everything that will ever be invented has already been.
no really. it has.
idiot
Duped again
Why is it that supposedly hard nosed tech types keep cleaving to these marketing mantras like "Moore's Law".
It's not a law cos it doesn't have a proof. Its not even a theorem because it isn't based on any prior established proofs. Moore himself called it a "trend" and its likely continuance for another decade a "prediction" but these days it is really just a marketing soundbite designed to promote the buying cycle.
As a statement it is tremendously inaccurate in its terms which relate to two variables in a multi-variable environment and it seems to assume "all other things being equal" without any real reason for making such an assumption - particularly as the original statement was made over 40 years ago and carried a caveat by the guy who actually said it.
I am constantly amazed at the ability of supposedly intelligent people to completely ignore actual evidence in favour of theologising their prejudices.
Optical network, embedded memory, Graphene, Nanoptube and TSV will replace silicone downsizing.
Silicone tech may run out of steam in perhaps 10 yrs' time. Other techs like inter-core optical networking, massive embedded fast memory (resistive or phase change memory) as well as the application of new materials (graphene and nanotube), and 3-D chips (TSV) , will replace silicone downsizing.
Re: Optical network, embedded memory, Graphene, Nanoptube and TSV will replace silicone downsizing.
"Silicone tech may run out of steam in perhaps 10 yrs' time."
Silicone is wubbery polymer !
Re: Optical network, embedded memory, Graphene, Nanoptube and TSV will replace silicone downsizing.
Attached is the news release from ibm for your question.
http://www-03.ibm.com/press/us/en/pressrelease/33115.wss
Moore's Law???
If it was Moore's Law then it would not have 10 years to go. It is Moore's observation.
No mention of Nanotechnology?
I'm concerned that for an esteemed futurologist he made no mention of nanotechnology.
I had thought that the technology for producing ICs would transition into molecular manufacturing, bringing about a new era of mechanical nanocomputers.
When built in 3D, we'd fit the most powerful supercomputers into the size of a speck of dust, without the frequency, overheating and current leakage issues that affect electronics at this scale. Or did Eric Drexler get it wrong?
Re: No mention of Nanotechnology?
Nanotechnology is that technology in the range of 100 nanometers to .1 nanometer. Semiconductors were there in the last millennium where were you?
Drexler predicted there would be robot assemblers building things an atom at a time using any material. He predicted that carbon atoms would be assembled into diamond for instance. He ignored DNA and life and the limits of energy in such reactions that drive the ancient nano assemblers in all life. After billions of years does anyone really think that if it was possible to somehow not spend 1000 degrees C of energy in forming a diamond carbon bond life would have found it and used it instead of calcium bones, or or cartilage?
Drexler also had no place for electromagnetics in his actual implementation predictions.
Drexler got it wrong. Drexler got us talking and thinking about Feynman's point and he deserves a lot of credit for publicizing the field. No core nanotech inventions though that I'm aware of have his name as principal inventor.
Moore's Wannabe Law of Device Physics
Gordon expressed a hope and goal, not a law of nature or of man. Ever since inventors have used their imagination to sustain the hope and goal and the economy of much of the world.
Spoiled brat's like the readers of this rotter's blog haven't a clue where the miracles that make their culture and life come from or in many cases how core and critical they are to their well being until they are denied. Then like the obscene reaction of lawmakers and courts to the loss of their crackberry fix a few years ago the addicts will do anything, break any code, blame any helpless group to get back their drug.
When the miracle of ongoing Moore's Law Invention ends, so does the love affair with ever more powerful, less expensive personal computing including iPads, iPhones and all their contemporaries, and successors.
So also does the preeminence of Silicon Valley, and the economy of the USA.
This topic is closed for new posts.
