Researchers Boast First Programmable Nanoprocessor

schliz writes "Harvard University researchers have assembled nanowires into tiny 'logic tiles' that can perform adder, subtractor, multiplexer, demultiplexer and clocked D-latch functions. While the 960-square-micrometre chips are not currently as dense as 32nm CMOS technology, the researchers say future versions could be up to 100 times more efficient than current electronics, and could yield low-power, application-specific 'nanocontrollers' for use in tiny embedded systems and biomedical devices."

Comments

Good for cellphones

[currently_awake]

This will be too expensive for anything but a handful of applications. Cellphones is the biggest, as many people use subsidized phones so they won't see the cost. Having a cellphone in your watch that lasts all week without charging will be very handy. Should also be useful for police bugs/trackers, to ensure they can run on solar power. Also useful for implanted devices like pacemakers, can run off pressure changes in your arteries so no battery to replace.

Re:Good for cellphones

[by (1706743)]

Having a cellphone in your watch that lasts all week without charging will be very handy.

Actually, I think it would be very wristy.

I'll be here all week!

Re:Good for cellphones

[Lumpy]

http://www.bigboxstore.com/cellphones/wrist-watch-phones

Already exist. Friend has one of these china phones and it does go several days. it simply sits on a stand that inductively charges it.

bluetooth headset always in is annoying though.

Re:Good for cellphones

[Anonymous Coward]

http://www.bigboxstore.com/cellphones/wrist-watch-phones

Already exist. Friend has one of these china phones and it does go several days. it simply sits on a stand that inductively charges it.

bluetooth headset always in is annoying though.

Bah! Dick Tracey had one in the 1930's and upgraded to a wristwatch video phone in the 1950's!

GaAs?

[uigrad_2000]

From TFA:

The tiles each contained 496 programmable transistors built from ten-nanometre-thick germanium wires, and were touted as a "world first" in complexity and function.

There are a lot of semiconductor types made with Germanium. GaAs is usually the most common. It is already an established method of making small transistors, although no one has managed to bring the costs down to the same as silicon.

You would think that an article that talks about some new technology would actually state what they were building upon, and what they changed to improve the process.

Re:GaAs?

[mangu]

From TFA:

The tiles each contained 496 programmable transistors built from ten-nanometre-thick germanium wires, and were touted as a "world first" in complexity and function.

You would think that an article that talks about some new technology would actually state what they were building upon, and what they changed to improve the process.

It's right there, in the sentence you quoted: "ten-nanometer-thick germanium wires"

What's new is the ten-nanometer wires, not using germanium.

Re:GaAs?

[funwithBSD]

What might be really interesting is if these are immune to Cosmic rays (space exploration application) or EMP.

Even if they are only as fast as current technology, that is a huge leap for space. The tend to use older tech that resists CR damage better or shield the hell out of things

Re:GaAs?

[MozeeToby]

Why would they be immune to cosmic rays? If a ray hits a wire, you're going to get a voltage spike, which can be interpreted as a 1 when there should be a 0 on the line. Doesn't matter what size it is or what it's made out of. Now, if they're really talking about 1/100 the power it opens up the possibility of multiple redundancy using less power than current systems. Have 10 duplicate copies of your processor all working the numbers and they take a vote on the answer and there's virtually zero chance of cosmic rays causing problems, and you'd still be using 1/10 the power of a conventional processor (in theory... if their ideas work out... and are mass producible... and conventional technology doesn't catch up in the meantime...)

Re:GaAs?

[ALeavitt]

Except for the fact that GaAs is gallium arsenide.

Re:GaAs?

[Anonymous Coward]

Use of Germanium is indeed not new. But don't confuse with the commonly used Gallium Arsenide (GaAs), different element. Although there are uses of Germanium on GaAs.

Re:GaAs?

[ColoradoAuthor]

Germanium, atomic number 31, is Ge. Gallium, atomic number 32, is Ga. The logic devices in the article are made from single germanium wires (admittedly with shells of "stuff" around them); in conventional GaAs devices, gold or perhaps aluminum-germanium alloy wires are used as interconnects between gates, which are made by stacking different semiconductor layers.

Re:GaAs?

[enslaved_robot_boy]

These are made with germanium and silicon. No gallium involved.

The expense of GaAs comes from the incredible cost of the wafers. Experiments like this which use individually grown nano-wires use an incredibly small amount of material for the semiconductor layer. If you could actually make these devices at a large scale with good enough properties for commercial use they would be incredibly cheap and you could make them on just about any substrate (glass.. paper?... skin!??... ) no wafer required.

Re:GaAs?

[Anonymous Coward]

Or just make them too small for the cosmic rays to see and attack.

Re:GaAs?

[uigrad_2000]

Yes, I realized this after hitting submit, but I actually wasn't as far off as you might think.

The Mars Exploration Rovers and several satellites use triple junction gallium arsenide on germanium cells, and germanium-on-insulator substrates are seen as a potential replacement for silicon on miniaturized chips, according to wikipedia. =)

Re:GaAs?

[Thing 1]

It's right there, in the sentence you quoted: "ten-nanometer-thick germanium wires"

What's new is the ten-nanometer wires, not using germanium.

Is this really germane to the discussion?

Opti-moo!

[Anonymous Coward]

and Ulti-moo!

Return to discrete components?

[dazedNconfuzed]

So...is this a return to the days of building computers from discrete components (separate basic electronic components wired together), just smaller?

Who's going to build the first nano-PDP8?

Re:Return to discrete components?

[enslaved_robot_boy]

This is cooler than that.

This a two dimensional array of transistors that can be turned on or off individually.

It's like the screen on your monitor except that each pixel is a variable transistor. So you can imagine that one "image" in the two dimensional array would be perfect for compiling code or something and another image would be best for graphics rendering. Essentially the promise of this design is that you could change your hardware as fast as you refresh your TV screen.

Re:Return to discrete components?

[SanityInAnarchy]

We already have FPGAs.

Rather than research it myself, I'm going to be lazy and just ask: How much better (or worse) is this than existing stuff like FPGAs?

Re:Return to discrete components?

[enslaved_robot_boy]

As a materials chemist the benefits of this design compared to existing programmable gate arrays is beyond my knowledge.

However, I can say that this is exciting because several of it's design elements are made with simple processing techniques that don't require the amazingly complex (and expensive) equipment used in traditional microfabrication.

The design (transistors made in arrays of crossed lines) is very interesting to material scientists because it is so simple compared to a traditional transitor array (or FPGA for that matter) that it is possible to imagine building it using self assembly techniques at very low cost.

As someone who works in nanotechnology, I can tell you that bottom-up transistors arrays is one of the coolest things that might actually be possible and that this is a good step in that direction. Essentially this research is a step towards being able to grow a computer chip instead of build one.

Re:Return to discrete components?

[jrobot]

This chip has 496 transistors, a modern FPGA has on the order of 1,000,000,000.

that vs quantum computing

[mapkinase]

Much less publicity noise on this and so much noise on quantum computing.

Re:that vs quantum computing

[Anonymous Coward]

Noise destroys quantum coherence, you insensitive clod!

Re:that vs quantum computing

[MozeeToby]

It's fair I think, this is really just an incremental step toward the local maxima that we've been approaching for 40 years now. Quantum computing, in theory, could be used to speed some calculations by several orders of magnitude, and that would just be the first working generation. Lets face it, if all we ever research is smaller transistors eventually we're going to get transistors to the minimum theoretical size (beyond which electron tunneling produces as much noise as the actual signal does). Granted, there's a lot of room for improvement after that point, with larger dies, better cooling, and faster clock speeds, but the underlying piece of technology (the transistor) will be virtually stagnant in 20 years.

Things like quantum computing, spintronics, 3D chip design, etc have the potential, if any of them work out, to break out of the local maxima and into a new part of the graph, maybe a part of the graph with a much, much higher peak. If you say you made a better transistor people go 'meh' because everyone knows that we can make better transistors, we do it every day. If you say you have a potentially paradigm shifting idea that may or may not work people get more exciting because... well, because the possibilities are more exciting quite frankly.

Obligatory

[bigredradio]

Yeah, but can I install Debian unstable on it?

Re:Obligatory

[Anonymous Coward]

And it'll run Crysis at full settings in Wine at 60fps.

Re:Obligatory

[SharpFang]

Yes, but it will be very -unstable. Like, it won't even start booting.

Re:Obligatory

[lennier]

Yes, but it will be very -unstable. Like, it won't even start booting.

And if it does, the universe may be destroyed in every way possible (and a few which are essentially impossible).

Re:oblig

[MobileTatsu-NJG]

'Oblig' is short for 'obligatus', which is latin for "this joke will never be funny."

Kinda like Minecraft?

[Anonymous Coward]

...tiny 'logic tiles'...

Kinda like this?

Re:oblig

[Suki I]

I for one welcome our new Borg overlords.

Isn't 'singularity' more appropriate? Is there a proper plural of the Singularity?

Transistors + Memristors + Bottom-Up

[enslaved_robot_boy]

This work is significant for two reasons

1) because it uses self assembled nano-wires as the semiconducting element of the transistor. This is an example of bottom-up processing.
2) because it uses a three layer gate oxide which can be altered by applying a high voltage, turning transistors on and off.

Bottom-up processing is extremely cool because it offers the future promise of being able to make electronics using the fundamental chemical properties of materials. The idea is that under the right conditions you can grow electronics without using super expensive top-down processing like deep-UV interference lithography.

The programmable nature of the transistor, which comes from the long lasting and reversible electrochemical changes that 6-9V applied between the gate and source generates (kinda like a memristor), means that if you make a square array of transistors and then you can address each transistor in the array individually, turning it on or off. This allows you to change the chip "hardware" on the fly. Which could be cool for programmers i guess....

Basically this is amazing work.

Re:oblig

[webmistressrachel]

If this was "Star Trek", which it obviously isn't, I am sure they would simply coin the term "singularities" and everyone would understand exactly what was meant.

Unfortunately, we're not there just yet, so that's not obvious to the parent poster... but it is to me, and it is to the Romulans, who use artificial singularities to power their warp engines.

The bad news is

[Anonymous Coward]

it must be programmed in nano-COBOL.

This would be perfect for my bitgrid

[ka9dgx]

My bitgrid idea is simple... an array of 4:4 Look up tables in a grid. This is just the technology needed to make it small and fast enough to do Exaflops. I look forward to this process scaling up.

Biomedical devices?

[schristo]

application-specific 'nanocontrollers' for use in tiny embedded systems and biomedical devices

Why do so many new technological advances announce their applicability for use in biomedical devices? I mean, is that not just another tiny embedded system, or is the intent to assure us that it is to be used for good, not evil?

Manufacturing

[Anonymous Coward]

Manufacturing has always been the challenge in finding a replacement for Silicon. The Si integrated circuit manufacturing process is very mature and very efficient. In order for a new technology to overtake it, they need to be able to do it in places other than a lab.