New Alternatives To Silicon May Increase Chip Speeds By Orders of Magnitude.

First time accepted submitter Consistent1 writes "A paywalled article in the "Nature Materials" journal describes the use of Magnetite to achieve ultra fast electronic switching, albeit, at the moment, only at extremely low temperatures. According to a story on Quartz, the team, led by Dr. Hermann Dürr from the Stanford Institute for Materials and Energy Sciences hopes 'to continue the experiment with materials that can operate at room temperature. One possibility is vanadium dioxide.' Chips utilizing this technology may operate at clock cycles thousands of times faster than the silicon-based chips used today."

Comments

Hummm...

[gagol]

I taught we already had gallium-arsenide transistors. The problem is cost as it is reserved for application where power enveloppe is very thin (earing aids) and switching speed is critical (telecom equipment).

Re:Hummm...

[jwinterm]

I taught we already had gallium-arsenide transistors. The problem is cost as it is reserved for application where power enveloppe is very thin (earing aids) and switching speed is critical (telecom equipment).

Another problem with GaAs and other III-V semiconductors is that they do not scale well, and so you can not pack as many transistors on a chip, and so they just can not compete with silicon in logic. They are quite useful for other applications, but not in your computer. Besides the low temperature hurdle, it's not clear if these new materials will face the same cost and scalability problems as III-Vs.

Re:Hummm...

[Seumas]

Did you . . . just call them . . . EARING AIDS . . . .?

Re:Hummm...

[dgatwood]

Yeah. They help you put earrings in.

Wait, what?

Re:Hummm...

[AdamHaun]

Wasn't it also hard to make decent P-type MOSFETS? I seem to remember that GaAs electron mobility is much higher than hole mobility, but it's been a long time since that one semiconductors class in college.

Re:Hummm...

[InvalidError]

One of the reasons silicon is great for mass-produced anything: silicon simply happens to be one of the most common and easily refined elements on Earth. For electronics, on top of being much cheaper than exotic materials, silicon's chemical properties (generally inert) makes it much easier to work with at high temperatures and caustic chemicals than most other materials.

Re:Hummm...

[Phase Shifter]

One of the reasons silicon is great for mass-produced anything: silicon simply happens to be one of the most common and easily refined elements on Earth.

The fact that pure silicon is an intrinsic semiconductor doesn't hurt, either. Just try making intrinsic GaAs...the amount of precision required to avoid making p-type or n-type material is ridiculous.

Re:Hummm...

[Agripa]

Besides not being suitable for complementary devices, GaAs also has no native oxide insulator.

Re:Hummm...

[RockDoctor]

Yeah. They help you put earrings in.

So, I need to get my step-daughter an EARING AID so that she can get her earrings in her ears not all over her face?

She will be so delighted by this news. I shall tell her immediately so that her reaction will happen in another country.

Overclock

[Azure Flash]

If this technology became mainstream, I'd bet my IBM Model M13 that people would still try to overclock the shit out of it.

Re:Overclock

[Anonymous Coward]

There would not be a lot of sense in a clocked design. If we are talking about a pico second switching speed, any signal would only travel about 0.3mm in that time. That really calls for clockless operation.

Clockless operation will likely converge faster at lower temperatures due to lower thermal noise, so the overclockers would focus their attention on undercooling.

Re:Overclock

[cupantae]

With a normal operating temperature of -190C, you'd probably need an extra fan or something to overclock it.

Re:Overclock

[tibit]

Thankfully there's a rather solid limit in how far that can go :)

Re:Overclock

[gmueckl]

Still, it's asymptotic. Have fun! ;)

Re:should be on the market in five years or less

[Anonymous Coward]

You do understand that somebody has to do groundwork before anything can be made in large scale. Even first silicon transistors where originally just proof of concepts until engineers where able to make manufacturing process around it.

Re:May...

[dinfinity]

Bet you were also expecting this one when you read the title: "albeit, at the moment, only at extremely low temperatures"
I know I was.

I thought latency was the main issue?

[Racemaniac]

I thought one of the main issues with increasing clockspeeds on processors besides heat is also the latency. at 3 Ghz a signal can only travel 10 cm anymore, and processors already have stages in their pipelines just to get the signals around. So going 1000 fasters would have to mean some major changes in how processors work i guess? since having your signal only travel 0.1 mm per clock pulse makes it rather hard to get the data around...

Re:I thought latency was the main issue?

[darkHanzz]

since having your signal only travel 0.1 mm per clock pulse makes it rather hard to get the data around...

There's still plenty of fixed-function hardware around (wlan chipsets, even though they're somewhat programmable) for which this might not be a major issue.

Re:I thought latency was the main issue?

[HetMes]

True, with conventional design the gain is questonable. But at least this would open up a new branch in chip design, and it would be interesting to see what comes out of it.

Re:I thought latency was the main issue?

[Racemaniac]

That's true, there are also chips that are meant for other purposed than computing, what bottlenecks do currently exists that current chipspeeds can't handle? You give the example of wlan chipsets, what would a faster chip improve for them?

Re:I thought latency was the main issue?

[MiG82au]

How about high speed and high gain amplifiers? Not everything revolves around digital logic.

Re:I thought latency was the main issue?

[serviscope_minor]

Latency is a problem certainly, but there's still some headroom. With a pipelined processor the signal doesn't have to propagate further than the next stage (ok that simplifies it a bit). At the moment, a top end processor is of order 1cm across (and now that's mostly cache and graphics), and even quite substantial ARM cores are down into the fairly small number of mm.

I suspect that unlike in the good old days, much like increasing transistor count no longer increases performance linearly, the same will go with clock speed once the processor is around one wavelength across.

One hypothetical way would be to have lots of really tiny, simple processors which are 0.01mm across, and then juice them up to 3THz.

Re:I thought latency was the main issue?

[stenvar]

I don't think that's such a big problem: you can still have large numbers of small processors that are extremely fast on local data but take a bit more to communicate with each other. There have been plenty of parallel machines like that already. Think Beowulf cluster, just on a much smaller scale.

Re:I thought latency was the main issue?

[Racemaniac]

i don't know anything about such chips, so care to explain why they would benefit from being faster? :)

Re:I thought latency was the main issue?

[swillden]

So going 1000 fasters would have to mean some major changes in how processors work i guess? since having your signal only travel 0.1 mm per clock pulse makes it rather hard to get the data around...

It seems like it would just change the design optimization criteria, making spatial distance dramatically between components dramatically more important than it is now. 3D chip design would become crucial, since it enables shorter paths. Of course, moving from flat or shallowly-layered designs to spherical construction would make heat dissipation an even bigger challenge than it is now, and would require completely new fabrication approaches.

Still "We have lots of really complex engineering problems to solve to make this work" is a better place to be than "Damn, we need to change the laws of physics".

Re:I thought latency was the main issue?

[rufty_tufty]

Many WLAN chipsets today use SDR(software defined radio), so most of the design is just a big DSP - so more clock speed = more complex algos. Alternatively since you'd likely have multiple channels in operation each of which probably has its own DSP by going faster you could put multiple channels onto a single DSP so save silicon area.
Or if you had hardened part of the algos into custom logic you could ease the memory latency requirements/move the hardened parts into DSP to save area.
Or move parts of the design that had to use onboard memories to use external memory to save area.

Lots of options and that's without me knowing the details of the design in question. As a general rule your 3 limits in a design like this are process speed, available area and external memory bandwidth; you're always at the limit for all 3 in any design, if you're not then you're wasting money

Re:I thought latency was the main issue?

[Sycraft-fu]

Ahh but remember the distance we are talking about isn't linear, but rather wire length. It is how far the electrons must travel through the pathways on the chip. That can wind up being larger.

Signal propagation is a real issue with high clock speeds. I'm not saying it is a kiss of death or anything, but it is something that can cause real issues with design.

Re:I thought latency was the main issue?

[Provocateur]

How many times i been telling you man, scour all the tech articles you want, research the hardware, have some nerd open up your case. You just really suck at Quake.

=)

Re:I thought latency was the main issue?

[Kjella]

True that, but cache latencies will have to go vastly up measured in clock cycles. If we say 3GHz = 10cm then 3THz = 0,1mm and an SO-DIMM module is 6.76cm across, you go from <1 cycle to 676 cycle latency just crossing the module. At those rates keeping the CPU fed with data might be the biggest challenge.

Re:I thought latency was the main issue?

[gagol]

Very VERY high frequency radio signal amplifiers? (radio telescope and all)

Re:I thought latency was the main issue?

[tibit]

Nope. The signal can travel as far as you wish, as evidenced by the DSN (deep space network) using the 8.5 and 32GHz bands at pretty significant distances within our Solar System. Voyager comms are in the 8.5GHz band IIRC.

The fact that the length of a clock pulse is physically small (on the order of 1mm) only makes it interesting from the engineering side of things, not impossible.

Re:I thought latency was the main issue?

[tibit]

So, the transputer is going to get a comeback? :) But seriously, transputers are alive and well. I'd salivate ever so slightly given an XMOS slice running at 1THz.

Re:I thought latency was the main issue?

[tibit]

I think only distributed transputer-style processing will be able to tackle that efficiently. Big networks of small CPUs with local memories will be "it". Assuming 0.2mmx0.2mm size of one compute-memory element, we'd have 4,000 such elements fit on a Haswell die.

Re:I thought latency was the main issue?

[darkHanzz]

For the current generation/standards at most some power efficiency. More processing power might allow better coding schemes, better beamforming (=less interference), smaller circuits (since less has to be done in parallel). So in the end it'll mostly come down to power efficiency. http://slashdot.org/comments.pl?sid=4025309&cid=44410675#

Re:I thought latency was the main issue?

[delt0r]

It has been suggested that this problem can be solved with asynchronous logic. An example of different signalling with very short switch times is with Rapid single flux quantum (RSFQ) logic.

Re:I thought latency was the main issue?

[Racemaniac]

if you read the entire comment, you would've noticed i meant per clockpulse, and that getting information around in processors at high frequencies is becoming a problem.

Re:I thought latency was the main issue?

[MiG82au]

Look up bandwidth-gain product and microwave electronics. Perhaps sometimes it's just better to not assume you know better than the researchers?

Re:I thought latency was the main issue?

[tibit]

It's only an engineering problem in the sense that pipelines are no more logical concepts, they have physical representation and you can't skip it. Those are of course solvable problems, only that the current CPU architectures aren't amenable to such treatment. That's not the end of the world, though, even now MS is pushing for parallelizing compilation.

Why FTL?

[amaurea]

Are you assuming that a signal needs to be able to propagate across a whole chip for each clock cycle? Otherwise, I don't see why the speed of light should be a problem here.

Re:Why FTL?

[Anonymous Coward]

The speed of light is actually a very important consideration, a signal can only move so far in a single cycle, if you operate at 1000 times faster you exponentially reduce the distance the signal can travel in that time and at a thousand times smaller distance you actually come into some very real physical limitations for the chip size and usefulness of the signal. It isn't that this has no uses, but it does have significant limitations on what this can be useful for.

Re:Why FTL?

[rufty_tufty]

No, the clock signal needs to time between two connecting flip flops nothing more. It's extremely common (i.e. it's about 5% of my job) to have to change the design in order to achieve this local clocking requirement.
That's without having multiple asynchronous clocks on a single chip.
Or asynchronous logic

Even when you need to do very long paths it's called a clock tree for a reason you can have a 1GHz clock that takes several ns to get from its source PLL to its destination flop because the delay through the tree to all the leaf nodes is matched. that is a 1ns period clock can take 4ns to get from the source to the destination, and that's all fine because as long as it's the same 4ns...
  Now things get harder when different bits of the chip have silicon that runs at different speeds so you can't balance the tree like you'd like to, but that's what makes this job interesting ;-)

Re:Why FTL?

[K. S. Kyosuke]

Also, a clock signal is a single-bit signal. You can use a wide interconnect for distributing it over large distances in the higher levels of the tree, making it much faster compared to the local interconnects. That makes it somewhat less of an issue than is the case with long-range data interconnects, which are parallel (or did they switch to serial lines even on-chip?), therefore have to use narrower interconnects, therefore are slower.

Re:Why FTL?

[Bengie]

Even the P4 had clock cycles short enough that they had to add stages in the pipeline to allow the signal to propagate across the chip.

Re:Why FTL?

[Rockoon]

If you want to have 1000 times shorter cycles, you need a 1000 times smaller chip.

Lets examine this..

The 80386 used a 1500 nanometer process. We are now playing with 22 nanometer parts (transistors that are 68 times smaller in length.)

The most common speed of the 80386 was 33 MHz, and the most common speed of a modern computer (according to the admittedly biased Valve Hardware Survey) is ~2500 MHz.

~2500 / 33 = ~75

So in practice what you are saying is clearly within an acceptable margin of true, but is perhaps not clearly stated (you need a 1000 times smaller process, not a 1000 times smaller chip!)

This does also show that the diminishing returns of higher clock speeds are likely real. If you want higher clock speeds without a smaller process size then you need a longer pipeline and thus higher instruction latencies, defeating a large chunk of the benefit of the higher clock speed.

However, for special purpose architectures (perhaps GPU's) with different use cases (where a deep pipeline doesnt have as many downsides), then higher clock speeds could be a big benefit even without a smaller process size.

Re:Why FTL?

[Noughmad]

So in practice what you are saying is clearly within an acceptable margin of true, but is perhaps not clearly stated (you need a 1000 times smaller process, not a 1000 times smaller chip!)

Well of course, it's a ballpark figure, not taking into account what exactly the pipeline does. However, it's not only the size of a transistor that matters. I don't know much about how a processor works (I am a Physicist), but as far as I know data in the form of electricity comes from somewhere (register), goes through some transistors, and then back into a register. If you want a cycle to last around 0.3 nanoseconds (which corresponds to 3.3GHz, close to modern i7's), the entire roundtrip can be at most 0.3ns / c = 9cm. So we still have a little room before the speed of light hits us with a brick wall, but with the current architecture we couldn't possibly have RAM speeds in the multi-gigahertz range.

Re:Why FTL?

[squizzar]

That's not really how it works... Propagation delay is not related to clocks (at least not in the way you seem to imply). With a stable and monotonous clock then you can easily propagate the clock to every point of the chip with a controllable delay (see comments from an actual designer of this stuff above). A simple clock tree, for example, can be implemented using a fractal like structure. Basically imagine a capital H, you have an equal length to each of the 'ends' of the lines from the centre. To each end you attach another 'H', half the size, with middle of the bar on the end of the previous larger H.

Your comments are referring more to the actual time between clocks that the chip has to perform some function. This is the propagation delay of actual logic. Again, the simple way to look at it is that each logic gate between the flops takes a fixed amount of time to respond to an input change, thus there is a limit to the number of logic gates in a chain that can sit between registers. There are other factors - e.g. setup and hold times for flops, clock skew (due to the distance between the register and an end point of the clock tree), fanout - but that's the basics of it. By adding pipeline stages you reduce the amount of logic gates that need to update between clock cycles, and thus you can run them faster. The downside is using more registers, more complexity, and in the case of processors particularly, the need to stall or flush pipelines, or predict behaviour of branches etc. so that the pipeline can be kept full.

Re:Why FTL?

[Valdrax]

[I]f you operate at 1000 times faster you exponentially reduce the distance the signal can travel in that time

Nitpick: Exponential does not just mean "extremely" when you use it with actual math and numbers. A constant multiplier is still a linear reduction, no matter how large or small.

Re:Why FTL?

[Rockoon]

Now if you make 1/1000 of 22 nanometers, you get 22 picometers. That's about 1/5 of the size of an atom. Good luck with structures of that size.

Yet atoms are not the fundamental building blocks of the universe... atoms are composites of smaller particles.

Re:Why FTL?

[garyebickford]

Some processors already use asynchronous signalling across the chip. In fact IIRC this has been a feature of CPU chips for some time.

Re:should be on the market in five years or less

[K. S. Kyosuke]

There's a huge spectrum between "the sample worked in the lab" and "we can ship complex CPUs to customers in million-sized batches". Sometimes it just turns out that a process is impractical. BiCMOS was dropped after Pentium Pro. Thermal output is becoming the bottleneck for Si these days, not switching speed. Also, whatever needs cryogenics simply won't end up in your desktop or cell phone.

Re: Dr. Hermann Dürr

[Anonymous Coward]

Fucking slashdot, with its lack of support for basic unicode. What is this? 1996?

Re:Too bad

[osu-neko]

Back in the days, when slashdot...

That's a bit of an obvious troll coming from someone with a seven digit UID... :p

Re:should be on the market in five years or less

[TheRaven64]

The first working Silicon transistor was 1954 and worked at room temperature. The first microprocessors were in the late '70s. It's great that people are working on other materials for transistors, but it's a very long road from 'works in the lab' to 'ships in a mobile phone'. 20 years is not unusual.

Re:should be on the market in five years or less

[lxs]

20-30 years seems to be a good rule of thumb. So if you want to know what the promising technologies of the next decade will be you should look at what has been done in the lab in the late '80s early '90s. (FDM 3D printing seems to be right on the mark, and if the Oculus Rift thing pans out VR will be too. Looking at stuff from the late '90s, electric cars will have to wait another decade to get mass adoption. LED lighting is ahead of schedule. Decent adoption rates a mere 20 years after the first superbright blue LED was demonstrated by Shuji Nakamura).

Re: Dr. Hermann Dürr

[Anonymous Coward]

1996? This is SLAAASHDOOOOOTTT!!!

*Kicks UTF-8 into a hole.*

Re:Sleepy

[Required Snark]

If you think it's useless, why are you reading it? If you were being consistent you wouldn't bother. Reading it and then complaining is even worse. You have literally invalidated you own opinion.

Why are you posting here? Why bother?

Do yourself and everyone else a favor. Go away and leave the rest of us alone. We're better off without you, The only person who enjoys your whining is you. Stop it now.

I thought that power is the main issue.

[Mask]

How much energy it takes to switch 0/1 states? What voltage? As I am not in the field, it would take me too much time to extract this information from the article (what is "trimeron annihilation" and how/does it relate the classical hole-electron recombination?).

I assume that it is possible to be 1000 faster only if it takes considerably less energy to switch states. It means that even if the latency constrains the speed, it would still produce less heat and will allow simpler clock/power lines.

As I understand it, one of the major factors that slow the speed of today's electronics is power. Be it in the form of routing constraints (possibly wider metal lines and possibly wider minimum distance between them), power dissipation, battery capacity in mobile devices, or cooling in servers, all are constrained by power. If this technology can lower power requirements then there will be a significant speed-up either in the form of more cores on a chip, or newer computation models that work better with deeper pipelining or with wider SIMD operations.

Another potential advantage of the fast switching is that it enables or enhances other computing models. Maybe we will move farther away from a pure CPU programming model to an FPGA/CPU hybrid programming. It's time to brush up your VHDL/Verilog capabilities, or to teach your pet language (compiler/interpreter/JIT) how to emit an efficient HDL. The advantage of FPGA programming is that you can define your own pipelines according to the computing task at hand. Another thing to consider is that with these switching-speeds it could be profitable to time-share an FPGA. Finally, it may be possible to reprogram an FPGA in less than a second.

Re:should be on the market in five years or less

[rasmusbr]

Electric cars were already off the shelf products in the 90's. The 1890's... That was before oil became really cheap.

The 30 year rule is a nice rule of thumb, but it's only a rule of thumb and it is subject to competition by other technologies.

Will it pan out?

[wbr1]

I seem to remember about 10 or so years ago a bit of talk about diamond semiconductors.

IIRC, making P-type material was easy doping with boron, and someone had finally come up with a way to make n-type material.

In addition, around that time there were two or three startups looking to manufacture diamonds using various -cheaper- processes. The combination of these things was supposes to give is diamond based chips that, due to the incredible heat resistance of diamond, could tolerate much more heat and hence higher clock cycles.

Does anyone know where this went?

Re:Will it pan out?

[gweihir]

Nowhere. Just as likely this will go nowhere.

Re:Will it pan out?

[slashmydots]

They probably realized that CPUs rarely fail because of high temps. The board around it fails because of high temps. So the diamond chip would kill its board.

Re:Will it pan out?

[overshoot]

Does anyone know where this went?

The usual destination for exotic semiconductors: no way to build a good gate dielectric or field dielectric. In other words, not manufacturable in volume.

Magnetite?

[Chillas]

Does this mean I should stop having my dwarves smelt it into iron bars?

Re:Magnetite?

[stewsters]

You can still use limonite or hematite i believe. You should also consider upgrading your facilities to produce something nicer.

Re:should be on the market in five years or less

[tibit]

BiCMOS is alive and well, thank you very much. It's just silly to use it for CPUs. Was it even used for any Intel chips at all? What for? It's pretty pointless unless you need bipolar-specific analog stuff on the same die.

Vanadium dioxide?

[jpallant]

The thing that immediately occurs to me is that this won't replace silicon. Silicon is massively available, it works, is well used and understood. Vanadium, in comparison is not. Plus, isn't it toxic? I know the semiconductor industry isn't what you would call green, but introducing an even more toxic element into the mix might not go down too well. I suspect this might, at the very best, have limited use in specialist applications. Making your computer thousands of times faster simply isn't going to happen.

Re:Think of all the opportunities!

[tibit]

Well, let's see. The Solar System weighs on the order of 10^30 kg. That's 2^100 kg. There's 2^86 atoms in a kilogram of hydrogen. That's only 2^186 hydrogens in our solar system, if its whole mass was hydrogen. You seem to be right - iterating through 2^256 is quite unfeasible.

Assuming iteration speed of 2^32/second, given 2^24 seconds per year, and a billion PCs worldwide (2^30), we could "crunch" only a space of 2^86. Our current resources are about a factor of 2^170 too small :)

Re:Too bad

[Anonymous Coward]

Back in the days, when slashdot...

That's a bit of an obvious troll coming from someone with a seven digit UID... :p

I've been reading Slashdot for over a dozen years, and I don't even have a UID because I never bothered signing up for an account. If I signed up now it'd be a very large number, and so would have a low perceived "seniority", and yet I remember when the Columbine and Hellmouth stories were posted here.

"may" == "will not"

[gweihir]

Ultra-fast circuits at very low temperatures are a very old thing: Josephson-circuits do it. That technology did not deliver, just as this one will not. Why the stupid headline?

Re:should be on the market in five years or less

[michelcolman]

Moore's law during 20-30 years doubling every 18 months makes a multiplication by 10000-1000000. Seems to be in the right ball park, then.

how is this going to work?

[ILongForDarkness]

At 3.5Ghz light travels 8.6cm per clock cycle. A thousand time performance improvement would mean ~86 micrometers. Ie roughly 400 transistor widths at current feature size. Since there are about a billion transistors in a chip assuming a square configuration you'd have ~31600 transistors on a side. Ie your 1000X chip would take ~75 cycles just to cross from one side of the CPU to the other. That is assuming speed of light which electrons definitely don't achieve. You still have to get electrons from RAM, disk drives, GPU etc. In short you'd need a massive pipeline to keep the CPU busy. The CPU might get 1000X faster but it will just be similar to (Amhdal's Law) parallelism given an infinite number of CPUs you are limited to the serial execution time, instead you are limited to the time needed to load and store back your program. Might get a 10X improvement with a 1000X faster CPU still nice but diminishing returns.

Re:how is this going to work?

[overshoot]

At 3.5Ghz light travels 8.6cm per clock cycle. A thousand time performance improvement would mean ~86 micrometers.

And have losses of more than 20 dB depending on the materials used for interconnect. Which means that the signal would have either be rebuffered every few microns or recovered at the receiver with something comparable to PCI Express but a thousand times faster.

Vanadium? I don't think so...

[Ralph Spoilsport]

Abundance of Vanadium: Earth's Crust/p.p.m.: 160

Abundance of Silicon: Earth's Crust/p.p.m.: 277100

Re:Vanadium? I don't think so...

[Sockatume]

That's not because vanadium is rare but because silicon is absurdly abundant; there's more vanadium than chlorine, lithium, cobalt, copper...

I really doubt scarcity is an issue here.

Re:Vanadium? I don't think so...

[Sockatume]

Oh man, that's a fun question. Some googling suggests that just wafer-grade silicon ingot production is in the tens of thousands of tonnes per year, per factory range. So it might be reaching a megatonne.

beware spectroscopists discussing electronics

[Goldsmith]

There is a reason we use different materials for high end optical and electrical switches. In material science we unfortunately see this all the time, where an optics group measures some interaction in a highly controlled environment and then projects that result onto a very complex electrical circuit. Generally optics groups which get published in places like Nature don't consider that they're measuring properties that are not actually relevant to a practical electrical circuit and not the only properties which might influence something like switching speed.

We could now step off an a wonderful discussion of rewarding over-reach in science, how the peer-review system is broken and how the publications-as-achievement system has derailed meaningful scientific advancement...

Re:should be on the market in five years or less

[Runaway1956]

I don't think the first PC's were produced in million-sized batches either. Lemme think a moment, and confer with my buddy, Google.

It is only AFTER some measure of success is established that lots of millions become routine.

http://en.wikipedia.org/wiki/TRS-80#History

French suggested that the company could sell 50,000 computers, but more skeptical executives disagreed and suggested 1,000 to 3,000 per year at the target $199 price. Roach persuaded Tandy to agree to build 3,500—the number of Radio Shack stores—so that each store could use a computer for inventory purposes if they did not sell.

Still forecasting 3,000 sales a year, the company sold over 10,000 TRS-80s Model Is in its first one and a half months of sales, and over 200,000 during the product's lifetime.[

That's nice

[overshoot]

Of course, most of the delay that limits clock speeds now is in the interconnect and not the switching devices. We're already using copper conductors and low-K dielectrics, so the next step is going to have to be superconducting interconnects.

Until then, it's mostly a laboratory curiousity.

Re:Think of all the opportunities!

[Thiez]

Yeah it's all rather disappointing really, but at least we know that gains in processor speed won't some day break our crypto (assuming it has no other weaknesses).

For some interesting numbers check out http://www.schneier.com/blog/archives/2009/09/the_doghouse_cr.html

Not Obvious

[b4upoo]

I strongly suspect that people are already suffering from future shock but have not put a finger on what is going on. Technology is a huge cause of job and social displacement at this time. It is not just the economy that is causing such chaos but the fact that less people can do a lot more work due to technology. Very fast and very smart computers will accelerate this pending upheaval. I am all for it but we need to be paying attention and doing triage on the wounded and displaced and even learn to identify changes that we do not tend to see clearly. One example is the tendency of youth to take wild risks lately. It is as if their future is hopeless therefore they feel they might as well destroy themselves. Radical stunts on skateboards are an example. For teens to seriously damage themselves over and over again and think that it is funny when they really do permanent injury to themselves is shocking. It even extends to risky drug use and teen suicides as well. They can not sum it up but in essence feel that they are worthless as they see no reasonable place in society for them in the future due to technology-computers-robotics making human labor less and less important. This has gone so fat that even education is questionable as a negative expense in that usually more education will not be enough to change anything in the lives of teens or young adults. We absolutely must get social policy to advance as quickly as technology. We really are not very far from replacing humans for truck and service driving. There go millions of jobs. We also are on the edge of fast food joints that no longer need humans in the cooking areas.

Re:Not Obvious

[KingMotley]

We also are on the edge of fast food joints that no longer need humans in the cooking areas.

Good, maybe one day we will consider pre-licked taco shells, or hamburgers with a splash of teenage junk a specialty item.

Re:Not Obvious

[iggymanz]

No, only slow economy causes job displacement. there is plenty of work to do that technology has created. in US the biggest increase in hiring is in "leisure and entertainment", followed by professional and business services, then retail.....this and housing market coming alive will soon ripple into manufacturing and construction

it's silly to be of the mindset, OMG, the lamp lighters and buggy whip makers and horseshoe smiths and chimney sweeps will starve!

Re:should be on the market in five years or less

[Anonymous Coward]

Times have changed though, and you have to make a big case for smaller batch sizes. Otherwise, a lot of the chip producers already have worked out exactly how many they need to make in what amount of time to have a reasonable chance of making a profit. Some friends who left academia for chip producing companies have complained of how often the tech they worked on got dropped from designs, all because it slowed down the process too much. This isn't like a factor of ten issue, but because more like they were given 30 seconds per wafer for one stage, and the new tech took 40 seconds instead, so it gets dropped.

For

[Impy the Impiuos Imp]

Isn't magnetite that natural iron form they make trinkets of to sell in Jamaican bazaars, typically in the form of animatable copluating humans, for placement as a dongle on mechanical security device unlocking portable storage ring?

Oh wait, that's its brother hematite.

Re:should be on the market in five years or less

[Bob the Super Hamste]

It wasn't cheap oil that killed those electric cars it was their range. Early on in the automotive world steam (external combustion), electric, and gasoline vehicles were all competitive but the technology for internal combustion engines progressed faster than it's competitors thus allowing greater benefits if using an internal combustion engine. Electric vehicles of started falling out of fashion due to recharge times and limited range (sounds familiar). Steam vehicles had the problem where they needed to warm up typically taking about 30 minutes to be ready to go as well as needing to have water replenished or have a condenser. I would also hardly custom built playthings for the rich off the shelf products since while there were were auto manufactures then it really was mostly one off builds. For perspective here is a little snippet from wikipedia on Oldsmobile:

In 1901, the company produced 425 cars, making it the first high-volume gasoline-powered automobile manufacturer.

I would hardly call 425 vehicles produced, not sold, in a year off the shelf and even then gasoline vehicles weren't what people were buying who could afford a horseless carriage (very apt description of these early vehicles) as electrics and steam vehicles outsold gasoline ones. Other than for collector or historical value you wouldn't want one of those 1890s era vehicles regardless of the power source as a daily driver.

Re:Too bad

[phantomfive]

Then check out this post and stop whining.

density/cost more important

[peter303]

Silicon replaced GaAs in the 1970s even though it was slower, because it could manufactured smaller for a much lower cost.

Re:density/cost more important

[iggymanz]

from the 70s to the 90s there also was reliability problem with massive transistor count GaAs chips, a battle Seymour Cray was fighting

Re: Dr. Hermann Dürr

[colinrichardday]

Hermann Dürr

What, is &uuml; so difficult?

Re:Think of all the opportunities!

[colinrichardday]

There's 2^86 atoms in a kilogram of hydrogen.

Hmm, there are 6x10^{23} atoms of hydrogen in a gram of hydrogen, so that would make it 6x10^{26} hydrogen atoms in a kilogram of hydrogen.

http://en.wikipedia.org/wiki/Avogadro's_number

Re: Dr. Hermann Dürr

[Anonymous Coward]

Maybe you should check your browser's configuration, or get a better browser. I can type "äöüÄÖÜß" quite fine, and it appears as intended.

Electromigration, we has it

[triffid_98]

We can already make silicon faster than we do, electromigration is why we don't. Switching to a different wafer material doesn't change the fact that we still have to interconnect the transistors somehow.

Bogus claim

[ChrisMaple]

The "1000 times faster than" current technology is blatantly false. They're claiming 1 ps. I couldn't find propagation delay data for the best current silicon processes, but 3 ps is a reasonable estimate, at room temperature.
They may have made a nice discovery, and it may be amenable to significant improvement, but so far they haven't demonstrated that they're going to replace silicon.

Re:should be on the market in five years or less

[K. S. Kyosuke]

Was it even used for any Intel chips at all?

Pentium and Pentium Pro. Dropped with the advent of Pentium II.

What for?

Execution units, I believe - but I'm not completely sure, I've read about that a long time ago.

Re:should be on the market in five years or less

[cheesybagel]

Another thing they did on internal combustion cars was to add the electric starter and electric lights. That removed the major advantage of going electric in the first place.

Re:Too bad

[Phase Shifter]

I've been reading Slashdot for over a dozen years, and I don't even have a UID because I never bothered signing up for an account. If I signed up now it'd be a very large number, and so would have a low perceived "seniority", and yet I remember when the Columbine and Hellmouth stories were posted here.

See, now we know you're faking it. If you could actually remember when the Columbine and Hellmouth stories were posted here, your nostalgia would be tainted by the memories of JonKatz articles.

Re:should be on the market in five years or less

[boristdog]

Yep, 20-30 years is a good estimate, especially since you also need to factor in the cost of the factory that will build something.

Regular silicon fabs using current feature sizes (and new toolsets) cost billions. Whereas older fabs with larger feature sizes (and older toolsets) that will still do the job for 90% of the applications needed can be picked up or built relatively cheaply in the hundreds of millions or even tens of millions.

Just because it can be done doesn't mean it will be done.

Re:should be on the market in five years or less

[garyebickford]

Just because it can be done doesn't mean it will be done.

Totally offtopic, but this immediately inspired the contrapositive:
"Just because it can't be done doesn't mean that it won't be done."
I'm sure this applies to something, somewhere! :D Politics comes to mind... :P

Re:Think of all the opportunities!

[colinrichardday]

Dang, I thought the GP said 2x10^{86}. My bad.