Moore’s Law Will End In 10 Years

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
Michio Kaku, professor of theoretical physics at CUNY, says Moore's Law will end in ten years. This is not a repeat from 1965, 1975, 1985, 1995 or 2005. Wait...I guess it is a repeat. :D

The two basic problems are heat and leakage. That’s the reason why the age of silicon will eventually come to a close. No one knows when, but as I mentioned we already now can see the slowing down of Moore’s Law, and in ten years it could flatten out completely. So what is the problem? The problem is that a Pentium chip today has a layer almost down to 20 atoms across, 20 atoms across. When that layer gets down to about 5 atoms across, it’s all over.
Click to expand...
 
oh wait chip stacking, layers, 3d logic circuits.. there's no way anyone can come up with other ways to increase power without 2 dimensional density increases.
 
He's got a point. The keyword in Moore's Law is "inexpensively". You can do 3D / chip stacking all you want but it's gonna be much more expensive to get more transistor density that way rather than a simple optical shrink. And even optical shrinks are getting relatively expensive (see current 28nm GPU prices).

Also, Kaku Kaku Kaku Kaku
 
At the current rate I predict in 10 years touchscreens will be so large you'll have to ram them with your whole body to produce an action.
 
for density, yes. from a performance standpoint it ended about 5 years ago. Now you get more cores and marginal improvements. Clock speed still < 4GHz. am disappoint.
 
Arcygenical said:
Pentium chip today.
Click to expand...
LOL

I wanted to post my thoughts on this since I work in the semiconductor industry, but I found a forum post which is the best single paragraph summary I've ever read on the subject.
To quote:

Negatron said:
If the limit is indeed atomic rather than subatomic then the estimate is somewhat correct, but far closer to 2018 than 2013.

Pitch/Gate length drops with a root of about 0.7 per full node reduction. So it's 32nm now, 22nm by 2011, then 16nm, ~11, ~8, ~5. 5nm gives you a distance of about 15-20 silicon atoms to work with, making each transistor consist of hundreds of atoms. This is believed plausible with lithographic technologies and silicon. 5 node reductions will take about 7-8 years.
Transistors other than silicon consisting of <20 atoms have been demonstrated. Lithography as we know it seems unfit for atomically precise manufacturing which will be necessary at that scale, but in principle it has been shown possible to construct such things. Processes 10 years from now may very well extended shrinkage for at least another few nodes.

When we do eventually hit the atomic limit, then the long road towards outward expansion with 3D circuits begins. This would make it possible to increase transistor count millions of times while negligibly increasing package size relative to a planar circuit.

Given a modern die size, it would be possible to ultimately increase transistor count a billion-fold over what we have today with atomic-scale processes. If it is possible by nature, given how consistent progress in integrated circuit manufacturing is, I think it should be assumed it is an achievable goal.

This could ultimately give at least ~Zettaflop performance on a single processor. After that there's really nowhere to go but out. I suspect transistor count will continue to drop as fast as always even after miniaturization limits have been hit. This is simply because manufacturing processes rather than material costs will continue to dominate the price for a very long time.

Whatever the ultimate limit is in transistor cost, it's so far away that when we finally hit it I doubt anyone will care. I'd be quite satisfied with yottaflop computer, that would make a rendering and physics engine as good as I could possibly care for.

Indeed, Moore's Law, if the rate persists, has decades of life left. Decades from now it may very well have decades more. Too early to say.
Click to expand...

I should emphasize that 3D stacking is already being used and supported commercially by some companies. This provides us an easy runway after we reach the practical limits of the existing lithography process. 3D stacking is almost totally unexploited today, so it is fertile ground for future improvements.
 
@Silverpike,

that's actually what he concludes. skip to the last paragraph
 
ahh, yes, the Heisenberg Uncertainty Principle. I was just writing a paper on this to the scientific community. ;)
 
"...but beyond that we may have to go to molecular computers and perhaps late in the 21st century quantum computers."

And the question will still be, can it play Crysis smoothly yet?
 
jiminator said:
for density, yes. from a performance standpoint it ended about 5 years ago. Now you get more cores and marginal improvements. Clock speed still < 4GHz. am disappoint.
Click to expand...

Yeah, because a 4GHz Sandy Bridge i5 with 3 of the 4 cores disabled performs roughly the same as a 4GHz Pentium 4. Maybe marginally faster. Uh huh. What are you smoking?
 
I love Micho's work and interviews but he's wrong about this.

They predicted some time ago that Moore's Law will end this year or that year. Quantum computing is the only technology in the lab that has the ability to "end" Moore's Law because it will simply supersede it beyond any known physics that could possibly process data faster.
 
nalc said:
Yeah, because a 4GHz Sandy Bridge i5 with 3 of the 4 cores disabled performs roughly the same as a 4GHz Pentium 4. Maybe marginally faster. Uh huh. What are you smoking?
Click to expand...
as far as I can tell, he's right in that performance hasn't doubled every year and half for quite some time
 
I think ol' Michio needs to stick to being the face for Discovery channel shows as the "standard Asian who knows everything"
 
I was thinking he meant that we are approaching the physical limits of putting traces on silicon.
 
far from slowing down things are speeding up if any thing we are a head of the curve
things might slow down for a bit but the they will jump back up

Singularity here we come :D
 
jiminator said:
for density, yes. from a performance standpoint it ended about 5 years ago. Now you get more cores and marginal improvements. Clock speed still < 4GHz. am disappoint.
Click to expand...
^This. Definition of Moore's "law" seems to bend like a noodle in a hurricane to suit current technology.
 
sleepeeg3 said:
^This. Definition of Moore's "law" seems to bend like a noodle in a hurricane to suit current technology.
Click to expand...

Well Moore's Law states that the number of transistors in an IC will double in 18 months, not performance, or cores, or whatever.

I agree that the definition has been corrupted.
 
you've pointed that distinction out twice now but you haven't explained why you think it's worthwhile to draw a distinction between the two claims
 
There's a biiiig difference between past claims of Moore's Law ending and current claims though.

In the past people just thought it would be too difficult/expensive to keep shrinking the sizes, whereas now we are quickly approaching the point where it becomes physically impossible to shrink them anymore. The size of a silicon atom isn't going to change so that is ultimately the limiting factor.

It's already becoming very difficult/expensive to keep shrinking the silicon manufacturing processes, just ask TSMC: http://www.extremetech.com/computin...y-with-tsmc-claims-22nm-essentially-worthless
 
travbrad said:
There's a biiiig difference between past claims of Moore's Law ending and current claims though.

In the past people just thought it would be too difficult/expensive to keep shrinking the sizes, whereas now we are quickly approaching the point where it becomes physically impossible to shrink them anymore. The size of a silicon atom isn't going to change so that is ultimately the limiting factor.

It's already becoming very difficult/expensive to keep shrinking the silicon manufacturing processes, just ask TSMC: http://www.extremetech.com/computin...y-with-tsmc-claims-22nm-essentially-worthless
Click to expand...

then we'll just switch to carbon. reason we haven't seen much progress in carbon based chips is because there's no need. Everyone is reaping the benefits of silicon. Moore's law isn't going anywhere anytime soon
 
Isn't there a physical limit to how small the gate can be before certain quantum phenomena takes over which cannot be prevented regardless of its material?
 
mope54 said:
you've pointed that distinction out twice now but you haven't explained why you think it's worthwhile to draw a distinction between the two claims
Click to expand...

Because that's what it states :p

Transistor density is something that's non-negotiable. It's a solid number; a fact. Performance, on the other hand, is dependent on the workload and is a general term. You can have a very dense chip that's a poor performer or you can have a very light chip that's not a good performer. Ultimately it all depends on the workload at hand. For example, Bulldozer is quite light at 1.2B transistors at 32nm and 315mm2 (light meaning not very dense compared to SB), yet it can still do quite well in threaded workloads. Moore's law never directly said anything about performance but it's something that people naturally ascribe to it regardless. It's not that it's wrong, because generally speaking it's not, but it's certainly not its actual meaning.

Moore's law will continue but the cost of shrinking will be an issue. Someone mentioned carbon but that would make a poor Si replacement.

There’s only one problem: Graphene isn’t really a semiconductor in the silicon/computer chip sense of the word. Unlike silicon (or germanium), graphene doesn’t have a bandgap, which makes it very hard to actually build a switching device — such as a transistor — out of it. Researchers have had some luck in introducing a bandgap, but graphene is still a long way away from being used in current silicon processes.
Click to expand...

http://www.extremetech.com/computin...er-silicon-that-could-beat-graphene-to-market

The transition to the Si replacement will come from modern innovations in Chemistry and our ability to produce the new substance for cheap rather than finding the best suitable replacement. Currently, anything "rare earth" might be considered a no-no if you've been reading the news.
 
Moore's law has already ended.

If you run the numbers, current gen CPU's fall short of his transistor count predictions.
 
Zarathustra[H];1038677758 said:
Moore's law has already ended.

If you run the numbers, current gen CPU's fall short of his transistor count predictions.
Click to expand...

To add more (moore? :p ) detail...

All Moore stated was that the number of transistors on a CPU will double every 1.5 to 2 years.

The first complete CPU on one chip was the 4bit Intel 4004 launched in late 1971. It had 2300 transistors.

If you follow that through and double the transistor count every 1.5 and 2 years respectively you find that more's law predicts that new CPU's launched today should have between 2.4 and 309 billion transistors.

Transistor counts of recent CPU's (different sources have different numbers so take these with a grain of salt)
Intel Sandy Bridge-E (6 cores): 2.27B transistors
AMD Llano: 1.45B transistors
Intel Ivy Bridge: 1.4B transistors
AMD Bulldozer (8 cores): 1.2B transistors
Intel Gulftown (6 cores): 1.17B transistors
Intel Sandy Bridge (4 cores) : 1.16B transistors
AMD Thuban (6 cores): 904M transistors
AMD Deneb (4 core): 758M transistors.

So yes, if you use the first single chip CPU as a starting point then we have fallen behind Moore's law. Only the Sandy Bridge-E is even close to the range predicted by Moore's law, and this is before figuring in that GPU's have been added on die.

Even Moore himself cautioned that the law would not be sustainable forever. Eventually you'd run up against the laws of physics and it would be more and more difficult to keep shrinking silicon. He predicted we'd have to move beyond traditional silicon for future designs, and when either happened (running up against physics, or switching away from traditional silicon) his law would no longer apply.

We simply don't need 10 more (moore? :p ) years.

The CPU with the highest transistor count right now, is already just shy of the low estimate based on it doubling every 2 years.
 
Isn't transistor count dependent on die size? I mean, you can look at Deneb vs. Llano and see that it has doubled (though I don't know the die sizes off the top of my head and I'm too lazy to search for them).

Then there's the other problem of what was the starting size of the first chip?

I'm not saying you're right or wrong, because quite frankly it doesn't matter if we abide by Moore's law or we don't [Moore's law not technically a law but more of a guideline as far as what to expect from transistor density in the future, really. IMO it should be called "Moore's goal" if we're being technical].
 
Azhar said:
Meh, the world's ending this year anyways.
Click to expand...
But it still won't run Crysis at max settings!!!!!!


That said, I'm not a computer geek guy really, but is there ANY sort of scientific basis behind Moore's law? or is it simply something some guy stated that for the most part has been somewhat true... due to the vagueness of the time frame.
 
edit: n/m found the answer with a quick google, it's basically a rule of thumb, and one that as mentioned has already failed (between the original Pentium and P4). Yet people cling to it and mention it like it's some absolute truth of the world.
 
sfsuphysics said:
But it still won't run Crysis at max settings!!!!!!


That said, I'm not a computer geek guy really, but is there ANY sort of scientific basis behind Moore's law? or is it simply something some guy stated that for the most part has been somewhat true... due to the vagueness of the time frame.
Click to expand...

It's based on the estimated guess of fab advancements. So, no. there's really no scientific basis. It's more of a prediction of the future.

Moore's this-better-adhere-to-my-rule-or-i'll-haunt-the-fuck-out-of-you law

Moore's i'm-taking-a-shot-in-the-dark-here law

Moore's oh-my-god-you-guys-actually-believed-me law

Moore's we-mustn't-disappoint-Moore law
 
pelo said:
Isn't transistor count dependent on die size? I mean, you can look at Deneb vs. Llano and see that it has doubled (though I don't know the die sizes off the top of my head and I'm too lazy to search for them).
Click to expand...

That probably has more to do with the fact that Llano has a GPU on die, and Deneb doesn't.
 
That still counts, though. It only says transistor density and technically the Llano APU's GPU side is a co-processor just as we considered the FPU many years ago. Moore's has-no-basis-on-any-real-scientific-theory law would still apply regardless of what you've got in the CPU/APU/SoC but rather addresses transistor density per-mm2 x time.
 
You must log in or register to reply here.
Back
Top