Thread: Moore's Law: Is There A Limit?

Yes and no. When you get down to atoms, or maybe earlier, you can't shrink the transistors any further. When this happens a paradigm shift will occur. In the case for processors, three dimensions will be used instead of just two. Because of this processors will keep on getting more efficient exponentially.

But once you reach a point after the three-dimensional change, wouldn't they have to get bigger, and for a certain size can only become so powerful? Granted the price will most likely continue to decrease, but once the size of the transistors gets down to an atom, then there would have to be a limit on how many can fit on a 3in x 3in x 3in cube? Sure it would have an amazing amount of memory (several yottabytes, in which one equals 10^24 bytes) but there would only be a change in the price, not the memory capacity.

I guess I just answered my own quesiton, but feel free to post any other theories.

Moore's law is really a small portion of a bigger law and that is the exponential growth of processing power. (I don't know if it has a fancy name.) So yes, Moore's law has a limit, but exponential growth of processing power does not have a limit. Today processors are being made more efficient by shrinking the size of the transistors. When that method loses momentum, transistors won't be shrunk anymore but a 3rd dimension will be used. This is really a very different way of increasing processing power (I don't know the details either). Our brain for example is so powerful because it is three dimensional.

Anyway, supposedly the law of exponential growth can be applied to a lot more than just processing power. You should check this guy out: http://www.ted.com/talks/ray_kurzwei...nsform_us.html

It's an interesting subject.

There is patently a limit, the only issue of contention is where it lies. The thing is every time somebody tries to put a cap on it, there's a paradigm shift and things are multiplied by a factor of thousands. We're getting quite close to the atomic limit but quantum computers represent a huge potential increase in computing power.

Quantum and 3d computers will raise the limit massively, but eventually you will only be able to have so much computing power in a set volume.

And once we exhaust the third dimension, we'll move on to the fourth spatial dimension... You heard it here first.

After listening to the guy in the link talking about projected technology, where is my computer in my shirt?

In both private and government owned industry Moore's law is of definite importance. There is a lot of progress being made along the lines of qubit processing via single molecule magnets in spin echo experiments, however there is still a ways to go until these things are actually integrated beyond the capability of a turing machine. My opinion is that during the time when quantum processors become a reality, that graph will look more like a logarithmic progression than anything due to a plateau effect caused by actual physical limitations of the molecules or individual atoms themselves, the smallest atom being hydrogen of course. In what way hydrogen atoms themselves might be used individually in qubit processing is a good question for any physicist today, since that is a seemingly ideal limit to reach. Here's some news from last year regarding this topic: First Electronic Quantum Processor Created

It's my understanding that quantum computers are a completely different ballpark though, and that the increase in power from molecular to quantum is far greater than the spacial scale factor?

I thought that each qubit could somehow interact with all of the others, meaning that the power you get with increasing the number of qubits increases exponentially, rather than linearly as in traditional computers?

I could be wrong though, and I have no good conceptual grasp of quantum physics (yet).

Originally Posted by Xei

It's my understanding that quantum computers are a completely different ballpark though, and that the increase in power from molecular to quantum is far greater than the spacial scale factor?

Generally speaking, we're talking about one and the same thing. Quantum processors are part and parcel of quantum computing, which, as I am sure you know, is engineering designed to take advantage of states like superposition and entanglement that produces viable ways of storing and manipulating information. I suppose a modified version of Moore's law could be applied to things like quantum processors, assuming they either get smaller or better.

The increase is productive power, in terms of the information being processed and stored, would absolutely be exponentially better than using, say, silicon because the spacial scale occupied by quantum mechanisms are drastically smaller, and can be carried out with increasing complexity almost instantaneously resulting in an increased efficiency.

Originally Posted by Xei

I thought that each qubit could somehow interact with all of the others, meaning that the power you get with increasing the number of qubits increases exponentially, rather than linearly as in traditional computers?

Yes, each qubit acts like a binary number, and one needs quite a lot of binary to create a good programming layer. With the huge quantity of qubits that can be utilized the difference in computing speed and power is enormous. The output produced depends entirely on the functionality of those qubits and how they're regulated to consistently produce and maintain, or "remember," spin states so that when you input something a desired output occurs, such as printing out a sheet of mathematical formulas at your home or continuously recording and graphically analyzing one hundred million particle collisions per hour at CERN and storinging them for later use. When we start talking about physical concepts like these it's good to recognize the dichotomous nature between the physics that is happening (energy, power, electricity, entanglement, superposition, spin, decoherence, etc.) and what that physic translates to, or the concept they contribute to (binary or qubit, programming, syntax, input, output, data, etc.).

Originally Posted by Xei

I could be wrong though, and I have no good conceptual grasp of quantum physics (yet).

I am a neophyte myself, no biggie.

I believe they're already starting to run into quantum-level limitations that are causing the curve to decrease in growth rate a bit, though there's plenty of creativity to go around still.

I'm a bit more concerned about I/O - that has NOT been following the Moore's Law trend so cleanly. We'll see what solid state drives get us over time, though.

Originally Posted by Replicon

I believe they're already starting to run into quantum-level limitations that are causing the curve to decrease in growth rate a bit,

Would you have any citations on that? I'd be genuinely interested to know.

The Dark Secret Of Hendrik Schon: Moore's Law

Originally Posted by ArcanumNoctis

The Dark Secret Of Hendrik Schon: Moore's Law

Excellent point, ArcanumNoctis. Knowingly, it is frustrating to me that I failed to expound that.

Originally Posted by Xei

It's my understanding that quantum computers are a completely different ballpark though, and that the increase in power from molecular to quantum is far greater than the spacial scale factor?

I thought that each qubit could somehow interact with all of the others, meaning that the power you get with increasing the number of qubits increases exponentially, rather than linearly as in traditional computers?

The current increase is exponential, not linear.

No, the increase in number of bits (per area) is exponential.

The increase in power associated with an increase in number of bits is a linear relationship.

Originally Posted by Phion

Would you have any citations on that? I'd be genuinely interested to know.

They've got a few years to go still, but once transistors reach a certain size, and are a certain distance from one another, tunneling begins to occur, which greatly reduces their reliability. Current leakage in a small field with millions of transistors, as you can imagine, will screw things up. Here's an article about it that google quickly brought up:

http://www.zdnet.com/news/intel-scie...res-law/133066

Though it looks like the cost of manufacturing will be increasing at a higher rate than the benefits of shrinking transistors, so maybe the real "Moore's Law Killer" will be economics before quantum physics:

http://news.cnet.com/8301-13924_3-10265373-64.html

CPUs hit their speed limit a few years ago, manufacturing will hit it's limit shortly. It will simply drive more intelligent designs, not stall the industry.

Originally Posted by Xei

No, the increase in number of bits (per area) is exponential.

The increase in power associated with an increase in number of bits is a linear relationship.

Ah, I misunderstood. You are right

Could quantum computers even entirely replace our computers today? As I understand it, measurement of a qubit is an entirely non-deterministic (although probabilistic) process, which would make it rather impractical for certain applications, wouldn't it? I imagine we will only, at least at first, see a kind of master-slave model, that is, a classical computer controlling a quantum chip and using it for certain algorithms and for solving certain problems.

I'm not holding my breath for quantum computing. A world-changing paradigm shift in physics is more likely to occur in our lifetimes than productionization of quantum computers in their current incarnations.

Originally Posted by illidan

Could quantum computers even entirely replace our computers today? As I understand it, measurement of a qubit is an entirely non-deterministic (although probabilistic) process, which would make it rather impractical for certain applications, wouldn't it? I imagine we will only, at least at first, see a kind of master-slave model, that is, a classical computer controlling a quantum chip and using it for certain algorithms and for solving certain problems.

Measuring Single Qubits

The Letter, titled “Signatures of Quantum Behavior in Single-Qubit Weak Measurements,” is authored by Korotkov and two colleagues at Penn State: Rusko Ruskov and Ari Mizel.

“The idea of this paper was to come up with something similar to Bell inequalities, but for solid-state systems,” Korotkov explains to PhysOrg.com. John Bell’s famous theorem, that predictions of quantum mechanics differ from what is intuited, and his famous paper, “On the Einstein Podolsky Rosen Paradox,” have formed the basis for deriving various inequalities regarding quantum mechanics. Many inequalities assume that projective measurements can be performed on the system. However, practically applying such measurements in a solid-state system is difficult. Korotkov and his colleagues suggest that rather than using projective measurements, Bell inequalities could be used to test quantum systems using weak continuous measurements.

Testing quantum systems is important when it comes to developing the next generations of computer technology: creating quantum computers. “Before attempting to construct a device for quantum computing, it is important to verify that a candidate system actually exhibits rudimentary quantum behavior,” Korotkov and his coauthors explain.

This is where Korotkov and his peers come in. Their system would possibly be a more practical way of distinguishing between a system that exhibits classical behaviors and quantum behaviors. The test proposed by Korotkov and his colleagues is one that makes use of quantum back action, a concept that many in quantum mechanics choose to ignore. “Such an experiment would prove that you can’t explain this principle in terms of classical physics.” Plus, using such a test would detect a system that “is required to evolve in the direction of the result, giving you information gradually.” And a system that does that, a system affected by observation, is an indicator of a quantum system.

While such an experiment detecting these signatures of quantum behavior is possibly three to five years down the road, another experiment, using the same underlying idea, has already been performed by an experimental group from UC Santa Barbara led by John Martinis, and in collaboration with Korotkov. Its results were published in the June 9th issue of the journal Science. “This experiment,” says Korotkov, “shows that the system remains coherent in a process of quantum collapse. This contradicts—or better to say—complements the concept of decoherence due to measurement, accepted in the solid-state community.”

He explains that decoherence is the norm for ensembles of quantum systems, but not for individual qubits. The recent experiment, though on a completely different system than proposed in the PRL paper, has the same basic idea: that even in a solid-state, “the state remains pure through the whole collapse process.” Additionally, “If you start with a mixed state, it becomes better and better known, which means purification of the state.”

The key, though, is continuous measurement instead of projective measurement. “There are already such experiments in optics, but this one uses a solid state system. It is completely new in that community,” says Korotkov. And it can lead to a practical way to test quantum systems and determine which would make good candidates for quantum computing. “Our PRL paper isn’t a breakthrough,” says Korotkov. “But it is a step in our correct understanding of quantum mechanics.”

by Miranda Marquit, Copyright 2006 PhysOrg.com. All rights reserved. This material may not be published, broadcast, rewritten or redistributed.

link

Thanks for the link. That all sounds very interesting. There's a lot that I don't understand, though.