Slashdot Mirror
HP's Crossbar Latch... Next-Gen Transistor?
moojin writes "CNN.com reports that "in a paper published in Tuesday's Journal of Applied Physics, HP said three members of its Quantum Science Research group propose and demonstrate a "crossbar latch," which provides the signal restoration and inversion required for general computing without the need for transistors.""
I think I speak for the sarcastic majority when I say this has little, if anything to do with Quantum Computing as defined thusfar.
I will be the first to admit that eventually there will be some limit to how small we can make a transistor (or transistor replacement) it seems that we still have a ways to go. I remember recently a spate of doomsayers going on about how circuits couldn't get much smaller than they are now, and how this would be the end of easy processor speedups. Well I guess they were wrong again. I don't think that will stop them from telling us that circuits can't be built smaller than this however.
Philosophy.
It'll be region coded. All the real power and functionality you want will be available in another region.
Life sucks and then you upgrade.
A feeling of having made the same mistake before: Deja Foobar
Don't get me wrong, this is great and all, (see a better article at EETimes) but to implement microprocessor-complexity devices with single nanometer technology, we need single nanometer scale wires and the technology with which to 'draw' them onto silicon.
We already have enough trouble at 90nm with wiring, and it's only getting worse at 65nm.
This looks like a great leap in device technology, but we need similar advances in lithography to really use it.
Well the first transistors had an ft (switching speed) of about 5khz and it didn't take much abuse to burn them out. Today we have transitors with ft's over 10ghz and they are quite rugged.
Once we figure out the basics, it is going to transform the way we computer simply out of the sheer computing power we'll be able to throw at things.
No, it won't.
Quantum computing (which has very little to do with the parent article) will change the way we think about computationally "hard" problems. Things like prime factorization, things like NP-completeness, things like cryptography.
But quantum computing will not replace the general-purpose Turing-complete model of computation we currently use. We will more likely see the idea of a quantum-coprocessor, something that you can interact with through a conventional CPU.
The problem with quantum computing involves the complexity of doing simple tasks... Yeah, it can factor absolutely mind-boggling numbers in one unit of time. It also takes that same one unit of time to figure out 1 + 1 = 2. The problem there involves the length of that unit of time - Between loading a state onto a set of qubits, them almost instantanously solving the problem, then reading the state off of them, you could have done potentially billions of cycles of normal CPU ops (no, I don't have a time-scale to quote for this, but I would consider it exceedingly optimistic to hope we eventually get it down to the millisecond level).
This development has so much potential because it points to a very, very major leap in the size of what we would currently consider a transistor... From 90nm, used by Intel and AMD's absolute latest mass-production facilities, down to a few nanometers. This means lower power requirements, faster CPU clocks, and much better areal density of functional units (getting down into the range of a few dozen atoms per switch, rather than hundreds of thousands at 90nm). The linked article also vaguely alludes to easier manufacturing techniques, but skimps on that one.
Overall, I consider your comment one of the most insightful I've ever read on Slashdot.
;-)
If they had cross-bar latch-based systems they wouldn't be have been using transistors in the first place.
We still use vacuum tubes and electromechanical relays alongside transistors. Perhaps crossbar latch technology simple can't handle large enough currents to interface well with the macroscopic world, so the aliens needed to use transistors to switch relatively massive currents up into the microamps...
So, no doubt in another 50 years, we'll find another layer of alien tech we have finally reached the manufacturing capability of making use of, and we can get down to using some cool property of the d orbital geometry as stressed in negative Scandium ions. No doubt the NSA's xenoassimilatory researchers missed this at the present time, since they considered it a mere impurity in the semiconductor substrate.
Sadly, I don't think it is. Browse around that site more, and you'll find all kinds of ridiculous conspiracy-theory crap. I like a good conspiracy theory as much as the next Slashdotter, but this site is just plain ridiculous, and it also seems outright contradictory. There's a big section there on how evolution is bunk and creationism is correct; that doesn't fit too well with the notion of UFOs and alien technology.
"...this thing is not a transistor... hence, end of moore's law."
If you look at the orignial paper, Moore is talking about "components," not specifically "transistors." There's no semantic reason why this couldn't continue to apply to the new technology.
The amount of computing that can be called "brute force" varies enourmosly between different problems.
For instance take the problem that Gauss is presumed to have solved in elementary school: add all numbers from 1 to 100. Gauss realized that the sequence can be broken into 50 pairs, 1 and 100, 2 and 99, etc, every which one adds to 101, so the answer is 101*50=5050. That's one example where we know one simple logical solution that takes much less computing effort than the more obvious addition of every number in the series.
OTOH, there are many other problems for which the only logical solution known requires a lot of computation. One well-known example is the four-color map theorem. Do you really believe that this theorem has been proved with "just" brute force, without the use of "science"?
I'm not an expert in the field, so I might be talking bullshit here, but as I understand it, there are problems in group theory where one can demonstrate formally that the only solution requires an amount of computation that's suspiciously close to "brute" force. The four-color map problem I mentioned required 1200 hours of computation when it was first solved in 1975.