Intel Experimenting With Nanotubes

illeism writes "C|Net is reporting on Intel's experimentation with nanotubes in processors. From the article: 'The chip giant has managed to create prototype interconnects — microscopic metallic wires inside of chips that link transistors ... Carbon nanotubes ... conduct electricity far better than metals. In fact, nanotubes exhibit what's called ballistic conductivity, which means that electrons are not scattered or impeded by obstacles.'"

Comments

Someone Alert Ted Stevens!

[vought]

Tubes are ascendant!

Truly, Ted is a technology genius. It's only a matter of time before these "nano tubes" are implemented to speed delivery of Internet content.

Re:Someone Alert Ted Stevens!

[Darth Liberus]

Haha, I was going to say "a processor is not a big truck you can dump something on... it's a series of tubes!"

Re:Someone Alert Ted Stevens!

[AKAImBatman]

I'd like to see you fit a CPU through those tubes when some other schmuck is streaming his video CPU. Why, I sent my 'mov 0x000010D0' CPU 300 milliseconds ago, and my CPU just ended 2 milliseconds ago! You have to understand that those tubes can be clogged! :P

Re:Someone Alert Ted Stevens!

[Sloppy]

Steven's wasn't as prescient as you think. Remember all those ads about Intel CPUs making your internet faster?

Nanotubes?

[JFMulder]

You mean like really really small Internets?

Quantum Dots

[googlebear]

Hey this is all really interesting stuff ...I think getting Intel behind some of the manufacturing technicalities is a major boon to the industry. Nanotubes, if intel's research confirms this, should prove to be useful in many different applications from mass power distribution to an elevator to the heavens.. who knows .. stay tunes.. also as an interesting side note.. VLSI will hit a rock bottom soon... I did a presentation in my Nanotechnology class last Spring on Quantum Dot Cellular Automata . This uses the electromagnetic repulsion of electons to propegate signals across molecules that are arranged in such a way to form logic gates.. http://www.nd.edu/~qcahome/ -Ian ian at ianroessle.org

3D Microprocessors

[AKAImBatman]

This sounds like it could be of particular use in 3D microprocessor technology. With the number of cores per die ramping up at incredible rates, we're starting to bump into latency issues again. I know that several memory manufacturers (who experinece similar die-space problems) have already switched to layered components to help relieve the issue and keep their dies smaller. But if we can weave nanotubes, we could do a lot more than just stack transistors three or four levels deep. Assuming that a inexpensive manufacturing process were developed, the chip could actually be fashioned in the shape of a cube. The result would make the chip orders of magnitude more dense than the CPUs of today!

Besides, it would look like a Borg cube under a microscope. How cool is that?!? :P

Re:3D Microprocessors

[rogtioko]

Another problem with stacked processors, besides heat, is that to really take advantage of the proximity the interface would have to be changed to one that integrates individual units of each processor more directly. This is far from conventional in terms of normal multicore-chip manufacture and would suffer from non-mainstream extra costs. Still, it should be designed and manufactured: the costs would go down when demand follows.
        I've read that, like 3d microprocessors, memory dies have often been stacked one on top of another (in slower DDR, DDR2 and NAND flash memory). The stacking allows good performance capacity upgrades with limited space; it's more cost effective! If Stacked memory dies sandwich a memory controller, the closer and faster operation would solve a big problem of distance latencies found in the cooperation of single memory dies embedded far apart on a flat circuit board with a memory controller. I could see a mainstream purely stacked memory chip. And if the nanotube interconnect idea works, it could be well implemented in both smaller individual dies and stacked ones.

Re:3D Microprocessors

[Iron Condor]

number of cores per die ramping up at incredible rates,

Yeah, we're already up to ... uh ... four...

Re:3D Microprocessors

[AKAImBatman]

GeForce 8: 128 Stream Processors
Sparc T1: 8 Cores w/4 threads (Maximum thoroughput: 32 simultaneous processes)
16 Core POWER5
Cell Processor: 1 Primary + 8 Sub-Processors

Intel Promises 80 cores

We're at a LOT more than "four".

What???

[Rellik66]

No Nanotrucks?

Power is Heat

[TheStonepedo]

If you get something running topped-out it may produce some waste heat. Thin chips with only a few layers can rely on a large, flat piece of some kind of substrate attached to a big heat sink and fan. If you make a cube-shaped processor, the innermost parts' heat will have to be dissipated through many other layers of working parts, creating a temperature gradient within the processor. If the innermost parts must be kept below a certain temperature, the outermost must be kept well below that temperature to allow for thermal conduction and the whole thing will have to run very cool relative to today's chips.

Re:Power is Heat

[AKAImBatman]

You are perfectly correct. This is currently one of the challenges facing 3D chip design. That said, there are several properties claimed by the article that would make nanotubes more suitable. First and foremost is that the nanotubes supposedly provide less resistance. Heat != Power per se, but rather the amount of resistance to the power applied. Less resistance == less heat. In addition, the amount of heat generated by resistance is also a function of distance. So the shorter distances provided by 3D nanotubing would provide less heat and overall power usage.

I'm also tempted to suggest that the empty space between tubings could be flooded with some sort of coolant to eliminate the temperature gradient; but I have my doubts about the feasibility of that. At such a small level, you'd have a lot of difficulty trying to fit atoms into that space. In addition, you'd probably do more to damage the circuitry than heat removal. Still, that doesn't place micro-heatpumps woven into the circuits entirely out of the question. Just mostly. ;)

In any case, we're already using WAY too much power to keep up these ridiculous clock speeds. Forcing chip-makers to scale the power usage back a bit wouldn't be all that bad of a thing. Especially if they're getting replacement speed increases from the smaller interconnects and lower resistance of the nanotubes.

Re:Power is Heat

[CODiNE]

Actually I remember an article here a while back about nanotubes being used to desalinize water. Apparently the perfectly smooth tubes aid the flow of water and defy the usual "size of pipe is proportional to water pressure" equations. What you could actually do in a 3D chip is leave extra nanotubes built in that simply flow in straight lines through the gaps in the chip where no conductive tubes are located, then pumping fluids through it wouldn't cause problems at all.

The excellent heat-transfer of nanotubes, plus the efficient water flow through them would make cooling them much better than current chips.

Re:Don't let random people write science articles

[Anonymous Coward]

People will conduct electricity (otherwise the electric chair wouldn't work), does that mean that people are made out of metal?

Re:Don't let random people write science articles

[wass]

What a stupid comment. If a carbon nanotube conducts electricity then it is by definition a metal.

Are you serious, or are you just trolling? As a blatant counterexample, there are non-metallic superconductors, which conduct electricity infinitely better than a metal. So sure, metals conduct (with non-zero resistance) and have some common characteristics, eg their fermi energy typically lies in the middle of a band (unlike semiconductors or insulators), ratio of thermal to electrical conductivity is relatively constant, etc.

But there are many things that also conduct fairly well at room temperature, such as doped silicon (an insulator). However, cool down silicon and the resistance increases (not enough thermal energy to excite electrons above the bandgap). Cool down a metal and its resistance will decrease (to a limiting factor). Cool down a superconductor and it undergoes a phase transition to a state of infinite conductivity.

Carbon nanotubes are actually extremely interesting in this regards, they can look metallic or insulating, depending on the chirality (ie, how the graphene plane is rolled into a tube). The metallic ones (with the fermi energy in the middle of a band) have quite long mean-free paths. Hence electrons can travel through the tube without scattering (this is the ballistic travel mentioned in the slashdot blurb). This limits the nanotubes resistance to the quantum resistance of about 25 kOhm. (Actually, the tube's resistance is 1/4 this resistance, as there are four quantum conducting channels because the graphene plane has two independent sites in its unit cell, and each site can have two values of electron spin).

Even some the insulating (or semiconducting) carbon nanotubes (or the graphene plane itself) are really cool. Due to the layout of the graphene plane, the band structure isn't pseudo-parabolic (as in a standard insulator) but conical (two cones meeting at a point), like a Minkowski light cone, or MCP from TRON. In the right orientations, the Fermi energy lies exactly at the intersection, and believe it or not, the excited states look EXACTLY like relativistic massive particles. The speed of light is mapped to the speed of sound instead, in this system. Really cool stuff, there are tons of future applications for nanotubes and graphene studies due to the interesting band structure, we've only really begun to break the surface.

Re:Don't let random people write science articles

[pyrote]

What a stupid comment. If a carbon nanotube conducts electricity then it is by definition a metal.
|People will conduct electricity (otherwise the electric chair wouldn't work), does that mean that people are made out of metal?

And if they float, they are witches! BUUURN..... er ahem.... carry on.

Re:Nanotubes good conductors of heat

[noigmn]

I'm not sure what is used in processors currently, but having the links as nanotubes would help the heat transfer within the material also. Nanotubes have a thermal conductivity of around 2000-3000 W/m/K at normal CPU operating temperatures. This is a huge increase when you compare it to the 149 W/m/K for silicon and 318 W/m/K for gold at room temperature.

So the increase in thermal conductivity by just having a proportion of the CPU made from nanotubes could possibly be enough to make up for the shape change. I wouldn't have thought much power would be saved by using nanotubes over any other conductor though. I'd be guessing most of the power loss is in the silicon gates, but I might be wrong.

http://www.pa.msu.edu/cmp/csc/ntproperties/thermal transport.html Carbon Nanotube Thermal Conductivity

http://en.wikipedia.org/wiki/Silicon Silicon Thermal Conductivity

http://en.wikipedia.org/wiki/Gold Gold Thermal Conductivity

Re:Don't let random people write science articles

[maraist]

Resistivity is the inverse of conductivity..
Conductivity is a function of:
A) Number of possible free electron states (positions) - function of temperature
B) Mean time to collision for given electron - function of temperature
C) [free] Electron density - function of temperature

Note that higher temperatures mean:
B) greater vibrational or translational properties of the material which obstruct the paths of free electrons.. So B is inversely proportional to temperature.
C) greater number of electrons are excited and thus broken from their atomic bonds (or at least lower atomic orbitals) and thus a greater number of electrons are available to participate in conduction. So C is directly proportional to temperature

Any material that
A) has free electrons (non-atomicly bound OR in covalent orbitals / orbitals shared between atoms)
B) has a non-infinite potential barrier between geometric positions
C) is above absolute zero

Naturally, the wider the path (radially), the greater the number of electrons and the greater the number of electron states, so the greater the conductivity.. It's 1 to 1 or linear growth.

As for length, the conductivity is an intrinsic measure, so length is somewhat irrelevant. However as a matter of practicality, on a large scale, the longer a path a given set of electrons have to travel, the more collisions will occur and thus the greater number of energy dispersals will occur and thus a greater amount of waste-heat. So you get an effectively greater measurement of resistance the longer the wire. This too is linear..

But length is usually a function of practical design (gotta connect two geographically distinct items).. Width, on the other hand is often a function of technology (how small can I make it) AND because width directly affects conductivity (gotta be wider to conduct more electrons), the intrinsic conductivity of the material dictates that for a given requirement of current and voltage, you must have a certain width for a given material.

However, not all geometries are created equally. The shape of the material, (which includes the curvature.. gentle curves v.s. right angles) affects the electro-static properties of the material. Indeed the mean-free-path to collision is different around the edges/boundries of a material, so obviously curved wire will have different properties than straight wire. Likewise the type of material adjacent to the conductor dramatically affects it's properties.

So to your original statement about water.

A) Water is a naturally polarized particle, so it can easily support attraction of free electrons.
    Ionized water (e.g. salt-water) has even greater electron attraction
B) Water is anamorphic (non structured, and constantly moving), so the mean-time-to-collision is pretty short.. This restricts conductivity significantly.
C) Water is not naturally ionized - it doesn't give off a free electron in it's natural state at room-temperature, so there are very few actual electrons available to conduct. Salt, on the other hand DOES give off a free electron when ionized in water. Likewise acids are ionized giving off free electrons. Thus lead-acid batteries use water with lots of free electrons, and thus conduct electricity reasonably well. Note that it's the storage of electrons, not the conductivity of electrons that makes these batteries useful. In fact the collisions due to high current conduction heats up the water.. This is how car batteries can explode, and this is why you should never open up the battery ports immediately after a car has been running for a long time.