Bye Bye Motherboards

[wacki]

Does anyone have anymore information on how this technology works, or how soon it might hit the market?


Linky

 

[Matthias99]

In the article:

The government agency wants to see U.S. manufacturers establish a new generation of supercomputers by 2010.

I would assume they'd be targeting that date if they wanted this to be part of that project.

However, I'm not seeing how wireless chip-to-chip transmission would be all-around better than wires. You have to deal with noise and interference (especially crosstalk if you have lots of these things near each other in the same frequency bands), you'll have to use some sort of analog-digital encoding (takes more time and power), it'll probably take more power for short connections without some breakthroughs in transmitter technology... it's a long way from happening as I see it, at least at high speeds and low power consumption.

 

[Ryoga]

That's my thinking as well. Wireless is for convenience and mobility. Generally speaking, wired communication is faster, more secure, and less prone to interference. I fail to understand why a wireless solution would be superior to using some kind of wired one. It's not like you can just pile all the chips together and have them work.

 

[wacki]

EM Field coupling doesn't have the same bottlenecks as RF data transmission. Different mechanism. Still I wan to read about this technology, cuz Im not EE.

 

[Scarpozzi]

Originally posted by: wacki
EM Field coupling doesn't have the same bottlenecks as RF data transmission. Different mechanism. Still I wan to read about this technology, cuz Im not EE.

I didn't have time to read the article, but that's what I was thinking. RF could never work for this technology.... I'm going to have to bookmark it.

 

[SonicIce]

Thats stupid. Why dont they just work toward room temperature superconductors?

 

[wacki]

Originally posted by: SonicIce
Thats stupid. Why dont they just work toward room temperature superconductors?

Superconducting materials solve a different problem. Superconductors solve the problem of resistance and energy loss. This technology solves the problem of capacitance and response time. Believe it or not, even though EM fields and signals travel very fast, the electrons themselves only travel 1 mm per second. Therefore, it takes time to charge a capacitor up. This is why you have to wait for a camera flash to power up. The larger the wire, the more electrons it takes to charge up a wire and send a 1 or "on" signal. This is why simply shrinking a CPU makes it run faster. This technology promises to decrease the number of electrons required to send a signal, and I'm interested in finding out more about it.

 

[CTho9305]

See links in my comment here.

 

[wacki]

Originally posted by: CTho9305
See links in my comment here.

Wow, thanks for the paper. The stats are very impressive. I can't wait for the technology to come out. Since they already have systems working at such impressive performance levels, I'm suprised they aren't selling high end systems with this technology.

 

[Calin]

Originally posted by: wacki

Originally posted by: SonicIce
Thats stupid. Why dont they just work toward room temperature superconductors?

Superconducting materials solve a different problem. Superconductors solve the problem of resistance and energy loss. This technology solves the problem of capacitance and response time. Believe it or not, even though EM fields and signals travel very fast, the electrons themselves only travel 1 mm per second. Therefore, it takes time to charge a capacitor up. This is why you have to wait for a camera flash to power up. The larger the wire, the more electrons it takes to charge up a wire and send a 1 or "on" signal. This is why simply shrinking a CPU makes it run faster. This technology promises to decrease the number of electrons required to send a signal, and I'm interested in finding out more about it.

You are wrong at several levels. Even if the electrons moves at just 1mm/s, there is no shortage of free (moving) electrons in the conductor materials. So even if a kilometer long wire will be traveled by a electron in 1 million seconds, the first electron to go out of the wire will go at the time the voltage was applied, plus the time needed for the electric field to travel the wire.
The long time for capacitors to charge is just related to the current capacity of the charger. You could have to wait some time to charge a big capacitor from a AA battery (or 8 AA batteries for 12V) than to charge it from a car's battery (the same 12V) just because the AA battery cannot sustain important currents.
The processors will run faster when the dimensions will shrink because the electric fields will have to travel shorter distances.

 

[TuxDave]

I don't think any of you understand what's being done in this research. They're not talking about wireless long-trace communications. They're talking very close distances. The paper cited is communicating over a distance of 126um or 126*10^-6 meters. The primary reason for looking into this is due to the parasitics caused by wirebonding chips. These fat, long wires are very inductive and you're forced to have many coupling capacitors to compensate. At least, that is my understanding.

Many other companies are exploring this 'wireless' i/o, and so far things look pretty good. I first started seeing papers on them at the 2002 ISSCC.

 

[Mday]

I dont consider that an end to motherboards. I consider it a new chip communications method which just uses coupling as a way to tx\rx. As I read more into it, alignment is going to be an interesting problem to deal with.