Nanowires Four Times Faster Than Silicon

evileyetmc writes "Advances in nanowires have shown that they may be the future in cheap, high-performance electronics. Researchers at Harvard have shown that nanowire transistors are are least four times faster than existing silicon ones. These nanowires show promise in being able to be embedded in plastics, and could lead to devices such as flexible displays that process information in the screen itself."

Re:Not ready for prime-time yet

[ggKimmieGal]

As time goes by, supply and demand will dictate the price of this new technology. Maybe ine 20 years from now it will be impossible to buy a silicon chip. At least, we can hope so. Nanotechnology has great potential to open the doors to inventions we cannot even begin to dream about. I suspect that over a very short time, these nanowire transistors will become even faster.

Still, there is hope for implanted computers.

If you mean computers implanted directly into my head... no thanks! Too creepy for me. Still, I know a whole bunch of people who are ready to get in line for that bit of surgery.

Nano-future

[UberMench]

I have talked with engineers at Tokyo University about this technology, and they are very confident that nanotube transistors are the future of electronics, not only because of speed, but also because they have fewer heat dissipation problems. And the prospect of having technology for electronic displays that can be rolled up like paper for easy transport just r0x0rz!!!

More Info

[Gridpoet]

There is a great in-depth article here

http://uw.physics.wisc.edu/~himpsel/wires.html [wisc.edu] ... very fascinating stuff the potential for small scale electronics is just staggering.

i wonder how long before they can mesh nanowires directly to nerve cells... plug me in!

Re:What about IBM's new transistor?

[RingDev]

The IBM chip is silicon at extremely low temperatures (~4.5K). This story is about carbon nanotubes being used in transistors. Two completely seperate and unrelated technologies.

-Rick

Re:What about IBM's new transistor?

[ChrisGilliard]

The difference is SiGe (Silicon Germanium) vs. Nanowire. The 500 gHz SiGe processor is something that can be made today. In fact it was made by IBM according to the article you linked to. The reason you don't see a commercial version probably has to do with the fact that it's expensive and consumes a lot of power. I would imagine it would be more economical to buy 500 1 gHz chips at $40 a piece (current bulk price for a 1 gZh chip). The nanowire chip has potential to be more economical. If we can learn how to incorporate them into current CMOS processes, they will be very useful because wires are actually one of the biggest components in chips believe it or not. These nanowires are so small (and apparently fast now too) that they'd make chips cheaper/faster/less power intensive.

nanowire != carbon nanotubes

[ChrisGilliard]

A nanowire is a wire of dimensions of the order of a nanometer (109 meters). They can be made out of Carbon Nanotube, but can also be made of other substances (e.g. Nickel, or Silicon)

Re:Isn't here more to it than 4x speed increase?

[Spellunk]

Your first paragraph seems right on, but the second doesn't seem to make sense. The length of the traces have little to do with speed, it is the actual switching speed of the transistor from off to on that causes the delay. Nano wires and transistors may switch faster, but the additional 10x improvement may come from heat/density savings, not the signal path length.

Re:Sorting problem.

[jnaujok]

The FA is about nanowires not nanotubes. They specifically point out the large difference between the two and say that nanowires can be made reliably and require no sorting. They also state that they are easy to make at room temperature.

What I find intriguing is that the article mentions how conducive nanowire technology is to three dimensional circuit construction with a per-layer size of 100nm. That means I can build 1100 layers into a 0.11 mm thick sandwich. How about 100 Athlon 64 CPUs intermixed with 1000 1GB memory arrays? With how reliable they are claiming this technology is, that would represent a 100 core CPU, with 1 Terabyte of memory mixed in. Seems like this is clearly the future of the CPU market. Especially if the heat disappation is as good as they claim.

How do you like my new Athlon 64 X100 with 1TB of memory running at 16 GHz?

Re:Wrong Conversion

[Anonymous Coward]

Why has this comment scored so highly? Breast implants are made of silicone. It's a very different material.

Sure they're faster....they're shorter.

[CFD339]

They're shorter. If you are talking about speeds measured at this kind of scale, the length of travel is a significant part of that speed gain. If you make the little electrons run further, they take longer to get here. The little bastards fairly sprint through the nanowires though.

Re:Isn't here more to it than 4x speed increase?

[Chrononium]

Yes, changing the length of the signal path does make a difference, but typically, computers process stuff at relatively low (compared to light) frequencies, implying that only a small performance benefit will be obtained by shortening signal paths. The biggest limitation to speed are the switching times of the transistors. Once you get down to the point where changing signal path length provides the best performance increase of all possible changes to the technology, you're hitting the physical (electromagnetic) limit of that material.

Re:Isn't here more to it than 4x speed increase?

[solitas]

Switching time (i.e. lead capacitance) is the worst bottleneck in chip operation: it can take a (relatively) long time to go from 'yup-that's-definitely-a-zero' to 'yup-that's-definitely-a-one'.

Re:What about IBM's new transistor?

[Lorkki]

I would imagine it would be more economical to buy 500 1 gHz chips at $40 a piece (current bulk price for a 1 gZh chip).

Unless you require a single chip running at 500 GHz for some specific signal processing application - in which case the complexity of the chip would not be that tremendous and the manufacturing costs therefore much lower. Not all ICs are meant to be general-purpose computers, after all. (Not to mention that actual processing power doesn't grow in a linear fashion as you add cores, but that's beside the point.)

You're probably right in that nanowires will have applicability in a broader range, and the embedded market will most certainly be thrilled to get their hands at them.