How Vacuum Tubes, New Technology Might Save Moore's Law 183
MojoKid (1002251) writes The transistor is one of the most profound innovations in all of human existence. First discovered in 1947, it has scaled like no advance in human history; we can pack billions of transistors into complicated processors smaller than your thumbnail. After decades of innovation, however, the transistor has faltered. Clock speeds stalled in 2005 and the 20nm process node is set to be more expensive than the 28nm node was for the first time ever. Now, researchers at NASA believe they may have discovered a way to kickstart transistors again — by using technology from the earliest days of computing: The vacuum tube. It turns out that when you shrink a Vacuum transistor to absolutely tiny dimensions, you can recover some of the benefits of a vacuum tube and dodge the negatives that characterized their usage. According to a report, vacuum transistors can draw electrons across the gate without needing a physical connection between them. Make the vacuum area small enough, and reduce the voltage sufficiently, and the field emission effect allows the transistor to fire electrons across the gap without containing enough energy to energize the helium inside the nominal "vacuum" transistor. According to researchers, they've managed to build a successful transistor operating at 460GHz — well into the so-called Terahertz Gap, which sits between microwaves and infrared energy.
Not a computing element (Score:5, Informative)
As a 450GHz computing element, this is a long way off. But it might lead to better terahertz radar. Right now, operating in the terahertz range is painfully difficult. It's a strange region where both electronics and optics work, but not easily. This may be a more effective way to work in that range.
Re:Not a computing element (Score:5, Informative)
That's mentioned in the IEEE Spectrum article (which by the way is about the most clearly written article on an early prototype technology that I've ever read).
The problems are:
-Too high voltage; can be mitigated by better geometry (probably).
-Insufficient simulations at present for improving the geometry, with the caveat that getting better performance (voltage-wise) might compromise durability.
-Because of the above, they don't have a good set of design rules to produce an integrated circuit. They're hopeful about this step, since the technique uses well established CMOS technology and there are many tools available.
Their next targets are things like gyroscopes and accelerometers. I'd say on the whole this strikes me as realistic and non-sensational. But if anybody knows better, I'd like to hear why.
Re: (Score:2)
Re: (Score:2)
It sounds like you're contradicting yourself - if accelerometers and gyroscopes already have a suitable housing, then the housing problem is largely solved - you could just as easily place a normal IC in the same housing. Though it would no doubt make heat dissipation a bit more of a challenge.
Pity Whitney Houston isn't still around (Score:4, Funny)
Stick her in front of a mike then tell her no more drugs and press record. That would have got you pretty close to that frequency range.
Re: (Score:3)
Yep - Mariah Carey hits the highest notes:
http://www.concerthotels.com/w... [concerthotels.com]
Whitney Houston is WAY down the list at #23, below even Elton John and Miley Cyrus.
Re: (Score:2)
From what I understand, drugs or not, her voice was already well past its peak (at best) if not totally f****d by the time she died anyway. Some of that was possibly due to age, most of it was probably smoking crack all day for years on end.
Speaking of drugs...
Re: (Score:2)
I am willing to predict that it will not happen as computing element. Computing elements have been limited by interconnect, clock distribution and the like for quite a while now. You cannot do longer traces in the GHz range, unless you spend an inordinate amount of chip-are for it. For analog, things are different, as you have few elements and there this tech may be interesting. But for digital, it is wholly irrelevant.
Re: (Score:2)
It might find a niche as a processor that makes money, literally, as in Bitcoin mining.
Re: (Score:2)
I think you may be confusing terms: low *frequency* optics behave as you describe, since low frequency = large wavelength. Low (short) wavelength on the other hand is the opposite, the optics need to be even more precise because the wavelength is even smaller than visible light.
Comment removed (Score:3)
Re: (Score:2)
Isn't this only a problem for branching? If you have a linear set of instructions to queue isn't it possible to start processing the next instruction while the previous is still propagating across the chip assuming the chip is laid out such that this kind of processing works and one instruction won't complete slower than the one following it. Sure it would have some downsides, branching would be expensive and out-of-order execution may be difficult but couldn't it in theory work?
Re: (Score:2)
It's not just branching. You also run into difficulties when a following instruction needs to use the results of the precending one.
Simultaneous multithreading (Score:2)
You also run into difficulties when a following instruction needs to use the results of the precending one.
If a scheduler foresees a pipeline bubble due to latency of the ALU, and data forwarding [wikipedia.org] is not enough to resolve it, the scheduler could feed the ALU a mix of instructions from two threads. This sort of simultaneous multithreading appears in Intel's Hyper-Threading Technology and AMD's "modules", and it's been around since the "barrel processor" architecture [wikipedia.org] of the I/O processor in the CDC 6000 mainframe.
2000 called and wanted its P4 back (Score:3)
Re: (Score:2)
Oblig: ... and it didn't even occur to you to warn *them* about 9/11 ? Sheesh. You could have saved us all a LOT of sorrow.
Re: (Score:2)
Re:The problem is not switch speed (Score:5, Interesting)
Asynchronous designs are faster (~3x) and consume less energy (~2x) but need an overhaul of the production process who is deemed too costly. Perhaps this technology could make it interesting again. (Source [columbia.edu])
Re:The problem is not switch speed (Score:5, Informative)
Not the production process so much as the design process. It'd mean starting over from scratch with a whole new architecture, redoing decades of work in hardware and software.
Re: (Score:2)
So? I would say that is bound to happen eventually anyhow. Traditional integrated circuits are quickly on their way to becoming a stick in the mud. Something fundamentally different will have to replace them eventually.
Re: (Score:2)
Re: (Score:2)
Not the production process so much as the design process. It'd mean starting over from scratch with a whole new architecture, redoing decades of work in hardware and software.
Presumably the hardware and software to which you're referring is the hardware to manufacture the chips and the software used to design them, considering that the asynchronous processor that was "faster (~3x) and consume less energy (~2x)" was an "asynchronous, Pentium-compatible test chip that ran three times as fast, on half the power, as its synchronous equivalent.", so the asynchronous processors themselves don't have to have a shiny new instruction set architecture. (The original PDP-10 KA10 processor [bitsavers.org]
Re: (Score:2)
but that with increasing clock speed the size of your chip is limited (as electricity can only travel that far in a given amount of time) -> can't keep your chip synchronized -> need to think of new ways how to sync everything / if there are alternatives.
I don't see why anything new is required. With today's design, bits are shifted from one section to the next on each clock pulse (or some multiple of the clock pulse, which just means that the internal clock is faster than the external clock).
Sure, the timing might have to be adjusted here and there. But you're still just shifting electrons short distances from one pulse to the next. If your chip die is 1" across, electrons can travel the whole width at about 10GHz. Since they seldom go a tiny fraction o
Re:The problem is not switch speed (Score:4, Informative)
One of the problems with increasing clock speed is gate capacitance and the RC time constant charging curve causing the switching FETs to operate in the linear region, causing power dissipation to go up with clock speed. This is why a decrease in process size has typically yielded a corresponding decrease in power dissipation at a given clock speed.
If you make the capacitance smaller, you can increase the switching speed (capacitance would decrease with the square of the feature size (gate capacitance is dependent upon gate area), wheras resistance would increase linearly, inversely proportional to feature width, assuming the feature depth doesn't change (resistance dependent upon cross-sectional area)).
Another poster has already mentioned asynchronous designs, so I'll pass on that particular nuance.
But clock propagation is a serious issue, and I can see a vacuum transistor improving this considerably.
Now, figuring out how far a wavefront will propagate in some period of time isn't too hard.
Undoped silicon has a relative permittivity of 11.68; the reciprocal of the square root of the relative permittivity is the velocity factor of a particular dielectric; for undoped silicon that's about 30% of c. Silicon dioxide, as used for most of the insulation on the typical MOSFET design, has a relative permittivity of 3.9 and thus a VF of about 51%. On a stripline laid on silicon dioxide (silica glass) the velocity of propagation is about 153 million meters per second, or 153 meters per microsecond or 153 millimeters per nanosecond or 153 microns per picosecond. 153 microns is a bit larger than the cladding on a typical fiber optic strand (most have a cladding diameter of 125 microns; OM1 multimode is 62.5 micron core/125 micron cladding, OM4 is 50 micron core/125 micron cladding, and single-mode is 8 micron core/125 micron cladding, for comparison). That's best case propagation time.
Now, to see how this translates to something of today, at least one of the models of the latest Haswell-DT Core i7 chips has a die size of 177 square millimeters. The chip is not square, and seems to be about a 4:1 rectangle in photos, which would yield about a 6.5 mm by 27.25mm die (yes, I know that gives 177.125 square millimeters; close enough).
Now, if a clock signal needs to go straight across the narrow portion, it will take about 42.5 picoseconds to do so, assuming transmission across silion dioxide alone. Propagation in the long direction would take about 178 picoseconds to do so, with the same assumption. The published top speed of this processor is at the time of this writing about 4.5GHz (I know it's a bit higher, but that's a moving target). This is a 222 picosecond clock period; easily doable in the short dimension, a bit more difficult in the long dimension, and probably already requiring some asynchronous elements and delay compensation. If you limit solely on clock propagation time, and are able to work in a slip of a full clock cycle, the long dimension will give you a limit of a bit over 5.5GHz; the short dimension will similarly give you a limit of 23.5GHz.
That's drastically oversimplified; each gate has it's own propagation delay that must be figured, and there are four cores (which makes it pretty understandable why the chip would have a 4:1 die dimension ratio, no?). A 20% clock delay factor will allow, with care, a good chance for synchronous operation (42.5 is pretty close to 20% of 222), but that's assuming straight clock traces (and they are not just straight across the chip).
Food for thought.
That would be handy for radio astronomy too (Score:5, Interesting)
Re: (Score:2)
Vacuum micro/nano-electronics are interesting for RF/mm-wave applications as the transport can be ballistic which could theoretically enable ultra-high-frequencies with scaling of the size.
I haven't yet found a paper for the 460GHz claim in the IEEE Spectrum article so I'm not sure exactly which figure of merit they have picked for that claim, but rest assured that their comparisons to other transistor technologies are highly flawed.
InP devices for example already operate up to 1THz power gain cutoff freque
Re:That would be handy for radio astronomy too (Score:5, Interesting)
I just noticed another disingenuous aspect to their claim - they say that because this operates at "atmospheric" pressure it will be more reliable than vacuum tubes of yore.
But these vacuum FETs are filled with 1 atmosphere of helium, so the partial pressure difference with the outside world for all other gases will still be the same as though it was operating with a full vacuum, and this device would require the same long-term hermetic packaging as a vacuum tube. It relies on helium to extend the mean free path of the electrons, though to be fair as dimensions are scaled down further from the current 100nm to say 20nm perhaps neither helium nor vacuum would be required. Still it seems to be a very misleading claim.
Re: (Score:2)
Re: (Score:2)
That's not how partial pressures work at all. It only matters about the relative pressure of each individual gas if you don't want it leaking in. If you want to keep out nitrogen and oxygen it makes no difference if the package has 1atm of helium only or a vacuum.
Re: (Score:2)
Even silicon certainly operates in the multi-hundred-GHz range, not the 40GHz which is for some reason cited in the article. Using graphene as a point of comparison is somewhat laughable as graphene has yet to demonstrate any truly practical advantage over group-IV or III-V transistor technologies, and has never been close to beating other leading device technologies on clock speed despite heavy press coverage.
Yeah the graphene comparison is spurious, except that it's a wider audience article and graphene
Re: (Score:2)
Yeah the graphene comparison is spurious, except that it's a wider audience article and graphene has been getting inexplicably large amounts of press recently.
A fair point, but I still don't excuse them for being part of the graphene press problem instead of the solution.
As for the other comparisons: what's the maximum speed of a MOSFET? You can get silicon BJTs into the hundreds of GHz, but I'm not sure about MOSFETS
Maximum published speed I've seen for a Si N-MOSFET is around 450GHz at 32nm, not sure of t
Magic Smoke (Score:5, Funny)
Re: (Score:3)
TFA says these transistors would be filled with helium gas, and if it gets replaced with other gases, the thing would quickly stop working due to ionization.
So I guess there'd be magic smoke going both in and out of the chip.
Valvistor? (Score:2)
Should this type of component be known as an "valvistor"?
Ahead of schedule. (Score:5, Interesting)
"It was a nice feeling to have a Microvac of your own and Jerrodd was glad he was part of his generation and no other. In his father's youth, the only computers had been tremendous machines taking up a hundred square miles of land. There was only one to a planet. Planetary ACs they were called. They had been growing in size steadily for a thousand years and then, all at once, came refinement. In place of transistors had come molecular valves so that even the largest Planetary AC could be put into a space only half the volume of a spaceship."
- Issac Asimov, The Last Question, 1956.
They will not (Score:2)
Today, Moore's law is an interconnect problem. The switching elements are pretty unimportant for it.
Ideal technology for high radiation/space (Score:4, Interesting)
This looks like the ideal technology for electronics that have to work in extremes of temperatures or high radiation environments. I'm surprised the military and aerospace industries aren't jumping all over this.
Re: (Score:2)
Gases ionize when hit with atomic decay particles/radiations. A transistor with a high probability of spontaneously turning on in radioactive environments sounds dangerous to me.
Discovered? (Score:4, Informative)
Natural things and phenomena are "discovered". Transistors were invented after a lot of hard work. By engineers.
Re: (Score:2)
I was going to give them the benefit of the doubt that they were talking about the phenomena using the end result as a convenient name. But no - that was apparently discovered between 1873-1884.
Well...even the first triode with a hard vacuum was back in 1915 (says Wikipedia with no citation). I'm thinking 1947 is merely the first commercial use of hard vacuum tubes on a wide consumer market scale. That's really the only thing I can see that lines up with 1947.
Re: (Score:2)
The 1947 is about the transistor, not the vacuum tube. The Bardeen-Brattain-Shockley transistor was developed in 1947, and got the three the 1956 Nobel in physics.
Re: (Score:2)
And yes - I was thinking triode and not transistor. I have no idea how I spent that much time writing without realizing that.
Re: (Score:3)
Why are we saving a law? (Score:5, Insightful)
Re: (Score:3)
Does that mean Obamacare isn't a law?
Re: (Score:2)
Moore's Law has never been a law and nobody treats it as one.
It started off as an observation which happened to basically be correct. Then it became more of a roadmap, with industry using it to set technology targets and allocating R&D resources so that they can continue following Moore's "Law".
Re: (Score:2)
'"Moore's Law" was simply a prediction.'
In the scientific sense, this is the only definition of a law: that it predicts.
I'm not holding my breath. (Score:2)
I'm still waiting for my memristor computer...
The first transistor (Score:2)
Julius Edgar Lilienfeld patented a FET in 1925. The FET is the type of transistor used in all modern CPUs.
More importantly... (Score:2)
...when will this result in a 100W Marshall head on a chip?
(Why yes, I am a guitar player! Thanks for asking.)
Re: (Score:2)
My *first* thought was, transistors were "discovered"? What, were they found nesting in a rock by a lake somewhere, pre-assembled? Nomenclature like this leads me to believe the submitter doesn't believe in any kind of IP whatsoever. They were invented. What was
I eagerly await ... (Score:2)
Re: (Score:2)
how about little nixie tubes on the CPU giving stats?
Re: (Score:2)
Can't get Nixie tubes anymore. All the Annie May fanbois have bought them up to build divergence meters.
How small can they actually make them? (Score:2)
Re: (Score:2)
The fad is ending.
Intended as a troll I think... but it's sadly true
Re: (Score:2)
Andy Warhol got it wrong / Fifteen minutes is too long.
Andy who?
Computers always were boring (Score:2)
Its what you can do with them thats interesting and thats only going to get more fascinating as the years go by.
A computer without a program is just a plastic brick.
Re:Planck trumps Moore (Score:4, Funny)
Just ask Madoka [wikipedia.org].
Re: (Score:2)
Moore's Law is about the cost to put a number of transistors in an area, not the size of the elements themselves. Generally we've reduced cost by reducing size, that is not necessarily the only way; you can continue the Moore's law curve by improving manufacturing in other ways until the cost of a chip completely covered in 4nm gates (probably the smallest the laws of physics will allow) falls to 0.
Re: (Score:2)
Fractal universe trumps Planck.
Heisenberg trumps fractal universe
Re:Planck trumps Moore (Score:5, Funny)
Re: (Score:2, Funny)
Gus Fring trumps Heisenberg.
Re: (Score:2)
Re: (Score:2)
Lighten up, people! Here, have a coffee, black with stevia.
Re: Planck trumps Moore (Score:2)
Now we have cat whisker problems instead of tin whiskers.
Re: (Score:2)
Re: (Score:2)
Too bad there is no such thing. Fractals are mathematical objects, not things that exist in reality except as approximations.
Re: (Score:2)
They are actual physical concepts, unlike fractals which are mathematical?
Re: (Score:2)
Presumably because Helium has some really interesting properties, many related to the fact that it's both a noble gas and has only a single electron shell. Which thanks to it's elongated shape and proximity to the nucleus allows for atomic behaviors and arrangements that are essentially impossible from any other element.
Re:Why is helium the new answer? (Score:4, Insightful)
They really are. The US government has been selling off reserves for below-production-cost for some time, causing prices to be artificially low.
Re:Why is helium the new answer? (Score:4, Funny)
Astrophysicists say no.
Re:Why is helium the new answer? (Score:4, Funny)
All we need to do is figure out how to mine the Sun and we'll have all the helium we could ever want.
Re: (Score:2)
If we could tap the sun, I guess we'd want something completely different first. And helium would actually be the waste product thereof.
Re: (Score:2)
What? Hydrogen? Jupiter is a *way* more convenient source for that. Both easier to harvest and far closer in terms of orbital energy.
Re: (Score:2)
How about power? If you can tap the sun, the first and foremost thing you'd go for is finding a way to siphon that insane amount of fusion power that gasball produces.
Re: (Score:2)
Well sure, and we're already doing so on a small scale - through solar, obviously, but also through almost every other energy source on the planet - solar energy being the ultimate source of most stabilized energy sources on the planet, depending on your reference timescale (biofuels, fossil fuels, even fissionables - though I suppose those are a product previous suns rather than our own)
But the *power* has no waste products - it's the *reaction* that has the waste products. Unless you're harvesting hydrog
Re: (Score:2)
didn't we reach peak helium some time ago?
Yes and no. Helium production is way down. Reason: shale gas. Helium is a by-product of natural gas production. "Tight" gas in shale formations, released by hydraulic fracturing, contains very little helium. But shale gas has driven the price of natural gas so low, that many of the conventional wells in Texas that contain significant quantities of helium have been capped off and idled.
So helium production is way down. But there is still plenty of helium in the ground, and it is available if/when the
Re: (Score:2)
The helium mafia is eager to sell.
The amount of helium likely to be used in these new semiconductors is infinitesimal compared to the amount used in airships, or even party balloons.
Re: (Score:2)
Actually, this one has been cropping up every few years for a long time. It has never delivered anything so far and there is no reason to think that it will do so this time.
It is time to get real: What we have in computing power in a "normal" chip these days is pretty much what we are going to get for the foreseeable future. That is not a problem. Software these days is so bloated and slow that there is a lot of optimization potential. And even afterwards, what do you want? Most things will work fine with c
Re:More expensive for whom? (Score:5, Insightful)
Intel has an insanely high Gross Profit Margin of 75%. That is the opposite of selling at a loss.
http://www.thestreet.com/story... [thestreet.com]
Re: (Score:3)
Re: (Score:2, Informative)
Actually making the chips is wildly cheap (always has been). They make a few thousand at a crack. It is all the other goop that goes along with chips that makes them expensive. If you read the original paper you will see Moore spends a good amount of time talking about packaging.
Testin, the package, pins, interconnects to the pins, wires to connect to other chips, the connectors, someone to glue it all together, etc...
The less chips you use the cheaper it is to make something. That is why moores law wor
Re: (Score:3)
the net profit margin of 15.12% trails the industry average.
Re: (Score:2)
People are dumb. Little details like these are wayyy beyond them. Or that Intel actually makes their profits somewhere else.
Re: (Score:2)
And Intel sells a lot more than CPUs and they do not tell anybody how they make their profits. Your argument is uninformed and worthless.
Re: (Score:2)
A) That is for the company as a whole.
B) That is not net profit.
Nothing you said actually contradicts my statement.
Re: (Score:2)
It's a good thing they don't have to spend almost all of that on R&D and facilities to manufacture newer tech in order to remain relevant.
Re: (Score:2)
It is possible for gross margin to be positive but net margin to be negative. In this scenario you are losing money on every unit you sell but you can make it up in volume.
Re: (Score:2)
Man, I wish I could sell at a loss with a 60% gross margin. Like all companies they make margins slim where competition is strong and large where it's weak or non-existant, but if you've ever had the impression Intel was dumping prices to squish AMD out of the market you must have lived in a different world than me. Dirty OEM tricks? Sure. Bleeding consumers dry by charging tons for extreme performance, long battery life or server features? Sure. Having superior process technology and pocketing the profit f
Re: (Score:2)
And you pulled this numbers for CPUs (!) from where? Right out of your ass? Because Intel does not publish these numbers and their net profit overall is far lower.
Re: (Score:2)
I certainly do hope so. They have been screwing their customers over long enough.
Re: (Score:2)
That's the benefit to owning the means of production and not giving a shit about what design they're churning out. Intel would like to beat them with Atom, but they don't have to.
Re: (Score:2)
Re: (Score:2)
keep weapons out of the civilian populations hands,
Huh? What?
I don't think the military is worrying about Joe Sixpack cobbling together a millimeter radar guided SAM in his garage.
Re: (Score:2)
The highest end Intel processors have 3-4 billion transistors.
Nice intentional troll (Score:2)
Sorry to say, it's too cohesive to be real. Better luck hooking the big kahuna next time, eh?
Re: (Score:2)
If you think that's bad, don't jump into the rabbit hole that is the Twitter feed linked from the web site:
Lastnight #emf massaging my brain and genitals all night, 3 #forced #ejaculation's from the #abuse: http://obamasweapon.com/ [obamasweapon.com] #rape #assault
Re: (Score:2)
Actually transistors were discovered then refined after experimenting with older technology.
http://en.wikipedia.org/wiki/T... [wikipedia.org]
>From November 17, 1947 to December 23, 1947, John Bardeen and Walter Brattain at AT&T's Bell Labs in the United States, performed experiments and observed that when two gold point contacts were applied to a crystal of germanium, a signal was produced with the output power greater than the input.[8] Solid State Physics Group leader William Shockley saw the potential in this, an
Re: (Score:2)
Where can I find a vacuum transistor tester. They took all of the tube testers out of the front of my local Radio Shack years ago, will they be replaced?
Yeah, and if you think it's a bitch to replace a burnt-out tube in your amp...