Slashdot Mirror
Future Phones May Use Vacuum Tube Chips As Silicon Hits Moore's Law Extremes (inverse.com)
An anonymous reader writes: A team of researchers want to replace transistors with vacuum tubes. Vacuum tubes are nothing new, however the ones in development at Caltech's Nanofabrication Group are a million times smaller than the ones in use 100 years ago. "Computer technologies seem to work in cycles," Alan Huang, a former electrical engineer for Bell Laboratories, told the New York Times. "Some of the same algorithms that were developed for the last generation can sometimes be used for the next generation." Dr. Axel Scherer, head of the Nanofabrication Group, said to the New York Times on Sunday, "Ten years ago, silicon transistors could meet all our demands. In the next decade, that will no longer be true." He argues silicon transistors can only take us so far. Vacuum tubes, for comparison, use tiny metal tubes that can control the flow of electricity. They're especially intriguing to researchers as they can provide a better solution to silicon transistors as they can consume less power and take-up a much smaller footprint. The report mentions they have the potential to bring an end to Moore's Law, even if silicon transistors show no signs of disappearing. For example, Lockheed Martin published new cooling methods in March that could help cool chips with tiny drops of water. With that said, Boeing has invested in researching vacuum tube chips. They may appear in the aviation industry before 2020, but it's unlikely we'll see Caltech's research appear in smartphones anytime soon.
I wonder if they are as resistant to EMP as old school tubes were.
Or is that simply a factor of size?
General Relativity: Space-time tells matter where to go; Matter tells space-time what shape to be.
Stop relying on devices that track you constantly and are closed for your computing. Stop using phones as computers.
Hmmmmmmmm. Or should I say "hummmmmmmm..."
(-1: Post disagrees with my already-settled worldview) is not a valid mod option.
"but it's unlikely we'll see Caltech's research appear in smartphones anytime soon."
I am reading on a phone right now you insensitive clod!
love is just extroverted narcissism
No, they wouldn't be as resistant on average, because yes, the biggest factor is size.
That being said, EMP resistance gets 'complicated', and it's easier to stick a small chip inside a faraday cage than a room sized monster.
I don't read AC A human right
Even siliconised 'tubes on a chip aren't new, exactly. TUDelft did it like last century.
The Fallout series had it right?
Call me a geek if you like, but I really enjoy watching this video of a guy hand-making triode valves (AKA vacuum tubes), it's somehow very therapeutic. Yep, only vaguely on topic, but what the hell, we're talking about vacuum tubes.
Oh no... it's the future.
Call me dumb if you want, but I design ASICs for a living. How am I supposed to design a chip with these devices. When I design in CMOS silicon, I have the choice of four different polysilicon well types (P, P+, N and N+). Do these devices require several voltage rails to provide bias, in the way that the dopant provides intrinsic bias in a FET?
I'm not old enough to have designed valve circuits, but from what I vaguely recall, you only get emission from cathodes, so with no hole mobility I don't understand quite how these things are supposed to provide complementary logic.
I've no doubt we can make ten or a hundred nano-vacuum tubes atom by atom. But compared to many billions of transistors? It looks like EUV litography @ 7nm will be ready by the end of the decade, but in the 2020s I suspect we'll hardly see any progress at all.
Live today, because you never know what tomorrow brings
Don't tell me, steam will also make a comeback.
That's gonna be so cool: switch it on and you hear:
Chug......chug...chug, chug, chug as puffy white smoke billows out.
And then Microsoft will tell you, "640 gallons of water oughtta be enough for anyone!"
Table-ized A.I.
or have they failed to miniaturize cranks yet?
if this is supposed to be a new economy, how come they still want my old fashioned money?
You're off by several orders of magnitude here. Think of the size of those 1970's vacuum tubes. Now think of the size of a single CPU. Now consider how much smaller said CPU already is. On top of that, consider the billion+ transistor count in current gen CPUs. So if the vacuum tubes are only 1 million times smaller, they have a long LONG way to go to reach the billion times smaller they need to achieve to even complete with current tech!
Computing is going to continue gaining power even if lithography stops. As some point they are going to figure out how to entagle quantum objects without having to cool them to 3K and the widespread quantum computer will be born. Quantum computers are like an analog and digital computer combined. There are multiple answers where digital only has 2, but the number of answers are limited to a finite set, unlike analog.
I also expect analog computing to make a reentry if for no other reason than AI. Brains are analog, and the reason we've had such a hard time with AI likely ties to this very phenomenon. We're using massive arrays of CPU's to approximate AI though big data analysis, but the computing power it uses is millions of times that of the human brain and the human brain will still beat it at most tasks. There is some very interesting research going on in analog computing.
Nope.
Just no. Nada. Not gonna happen.
And who at Slashdot took this seriously?
Bruce Perens.
with vacuum tubes.
So wait: You're telling me that soon the Internet WILL BE made up of a series of tubes?
He was a visionary, ahead of his time. (cough, just another clueless manager, cough.)
I find the "consume less power" claim a but surprising, given that vacuum tubes work by heating a piece of metal to white hot until it starts flinging off electrons. Sure, they're talking about making them very small, but the Apple A8 processor in my smartphone has 2 billion transistors. The heat from that many tiny vacuum tubes would add up.
Almost all materials including the new technology require electron energy change of order 1 ev to flip the state. So the power required will be limited by this. You need about few thousand electron flips for every flop. This will limit energy needed for each flop to be about femto-Joule or power efficiency of about 1 petaflops/w. This is still 1000s of times higher than modern computers but it tells that Moore's law is unsustainable beyond a couple a decades using classical computers.
I find that the results from numerical computations on today's transistor-based CPUs often have an undesirable "harshness".
Vacuum tube CPUs will hopefully yield richer, more mellow computational results.
My next phone will have the latest cathode ray display technology.
Some mornings it's hardly worth chewing through the restraints to get out of bed.
I wish more tech journalism acknowledged that by and large we are finding new ways to use existing tech rather than making it sound like startups have invented time travel. Anything that runs code is actually pretty milquetoast under the hood.
You know how Apple released the original iPad in 2010? And then they were hiring 10s of thousands of employees and spending billions on R&D without anything significant to show for it besides minor annual hardware revisions? Pretty much until Apple watch was announced in '15?
There are many reasons why hardware lifecycles are different from software. No idea about the real story, but what is released or not released in 2 1/2 years is not much of indication of anything.
Will this finally make tube-based guitar amps more affordable?
It's sad when you can buy a 150W solid-state amp with 5 DSP's that fairly convincingly models the behavior of a dozen tube amps for $300 but a 15W tube amp can still run you $1,000 easy and will be a total 1-trick pony.
Does anybody know how big they were 100 years ago? I have no idea. I'm guessing most people don't. Since when did "fraction of size of a vacuum tube from 1916" become a unit of length?
Seriously, how big are they? Assuming a vacuum tube in use in 1916 was 10cm in length, I'm coming up with 100nm, which is FREAKING HUGE compared to present day Si transistor sizes which are closer to 16nm IIRC.
This quote was amazingly stupid: "At this level, silicon starts to behave weirdly. It becomes more elastic, and starts to give out light. Silicon transistors also leak electrons at smaller sizes." It's not behaving "weirdly" this is purely a consequence of size. When things are small the ratio of surface to bulk is higher. When things are nanoscale there is almost no bulk left, so the properties begin to resemble surface properties more than bulk properties. Seriously, there's no mystery to it. I have a PhD in chemistry and have studied both nanomaterials and semiconductors. I've never seen such a stupid explanation of size-dependent properties as was offered in this article. I hope the Cal-Tech researchers didn't write that. Also, electrons don't "leak." That's just stupid. Current leaks. If electrons "leaked" you'd wind up with a charge that would oppose further leaking and the leak would stop itself. They phenomenon they are attempting to describe (and failing miserably) is leakage current. Leakage current happens when charge flows through an insulated path (i.e. the current going to the gate in a MOSFET). Was the concept too hard to explain simply while also being truthful? I think my explanation was fine, and it was just one sentence.
Anyway, I've seen some pretty bad writing before, but this was an entirely new level. There is nothing but speculation and horribly written incorrect statements about present day semiconductors. I would have dismissed this had they answered any remotely interesting question such as: What is the new advancement that enabled tiny vacuum tubes? How big are they? What are their electronic properties? How to they work? Why do their properties not change when miniaturized? Terrible story. The researchers at Cal-Tech should be ashamed if they had any part in this.
A vacuum tube is a macroscopic device. An electrode is heated, electrons shoot out and their trajectory is controlled by charged grids.
On microchip scales, it's all about quantum physics. Electrons are wave-like, they tend to teleport through obstacles, change size as they are heated or cooled down, really weird stuff. The math probably works but I wouldn't call these things "vacuum tubes" when the very notion of everything that makes up a vacuum tube is challenged at these scales.
"put an end to Moore's law"???
Don't the author mean an end to the end of Moore's law?
Unless of course he/she means that processing power would go beyond doubling every 18 months with this tech, since we're always on the cusp of breaking Moore's law downwards... Only the end never materialize, a little bit like doomsday announcers.
My phone is a series of tubes.
"a better solution to silicon transistors"
It's "a better solution THAN silicion transistors", you American cretin...
You don't say "We chose roses over daffidols because red is better TO yellow", do you?
I remember back in the early and mid-90's the idea was to include micron-size vacuum tubes in integrated circuits in order to cut down the reverse-current to a level that semiconductor diodes could not do it.
Those tubes were functionally only diodes and the idea was to use the strong electric field at the tip of a very small cone to achieve cold electron emission. Imagine hollowing out a half-sphere and then add a cone with the tip at the center of the sphere. Now apply a voltage between the two. The electric field is very strong at the tip of the cone, but much weaker at the interior surface of the sphere. This results in cold electron emission from the tip of the cone and a current when the voltage is applied one way, but no current when the voltage is applied the other way.
I have never seen these used in practice. I believe one reason is that such a component had to be at least a couple of microns across, and the chips had little use for a 2000 nm diode.
Anyway, such vacuum tubes could have some use in parts where zero reverse current is important, but due to size limitation don't expect to see them replacing transistors that are counted in the billions in modern chips.
And now a Marshall amp inside your Gibson guitar! Sweeeet.
Back when I was a young CS student, my teachers used to regale me of tales when there were people with full-time jobs riding around the computer (supposedly on unicycles) with a backback full of vacuum tubes replacing them as they burned out.
That's going to be a hellova interesting job now that their size is measured in nano-meters and there are billions of them on a chip.
Actually, "better to" sounds awfully British to me, but then I am just an American cretin.
I just assumed that it was a stupid mistake missed by a failed editor, possibly caused by the writer switching sentence structure in the last revision.
No, we chose roses over daffodils because red is PREFERABLE TO YELLOW.
But you failed English, apparently.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
Depends upon the context. If it's a noun, then "better than" is correct since it's a comparison with another object; if it's a verb, then "better to", e.g. "better to x y z". There's no difference between British and American English here which I'm aware of.
Relevant to portable devices? Yes. How vulnerable? Complicated.
You see, with an EMP it's typically a single pulse. You don't actually need a loop. The emp will 'shove' electrons one way or another, creating a voltage spike over [i]any[/i] wire of sufficient length along the wave of travel of the event. IE a line running perpendicular to the event would end up being a circle but not experience any significant voltage(0 for a theoretical superconductor of zero width), but one running from the epicenter out would experience maximum voltage.
What happens is that once the event passes the electrons reverse, and create a rush of current the opposite way. So if your diodes happened to be facing the 'right' way to simply pass the current, they might smoke on the 'backstroke' as the voltage overwhelms them.
Anyways, back to portable devices - while long runs like power lines are indeed a major concern, part of the 'problem' is that devices hooked up to the grid are designed for fairly dirty power(which an EMP spike is a form of), and hundreds of volts. Parts in many small portable devices are designed for 1-3 Volts. The problem is that even as our devices shrink, shrinking the runs that create the voltage, so hasn't their tolerance for said spikes - so the vulnerability remains, because today a run of mere centimeters can be enough to create a damaging amount of voltage/current on the board, or even in a chip itself.
For something like a cellphone, the antenna is also a concern.
I don't read AC A human right
... humming that song from the musical "On Broadway" while ruminating on the history of technology.
Back in the 1950s Bell Labs had the transistor as the latest and greatest new thing. General Electric was also working on miniaturized electronic devices with arrayed vacuum tubes. The G.E. idea was to build miniature vacuum tube circuits in a sealed housing that would be pumped to a vacuum to work. For repair or to modify circuits you would break the vacuum, open the housing, and work on it with jeweler's tools. After repair you seal the housing and pump it back to a vacuum. Bell labs won the contract for government money and the vacuum board technology dropped by the wayside in favor of solid state electronics. (anecdotal tale from the 1970s. A manufacturing jeweler my company did business with said that he had learned his trade working for G.E. on miniature vacuum circuits decades before.)
I've wondered if miniature vacuum tube circuits might not provide longer functional lifespans for satellites. Semi-conductors degrade in high radiation fields lowering the effective lifetime of equipment. (A ccd camera that would last years normally has a lifespan of only a few months in a >5 Rad/hr field)
NRRPT/RCT