Is this the silicon chip KILLER? Boffins boot up carbon-nanotube CPU

Look out, silicon: boffins in the US have created a simple working microprocessor out of carbon nanotubes, possibly paving the way for faster and less power-hungry computer chips. Nicknamed Cedric by its creators at Stanford University, it is the first of its kind: it can fetch software written in a subset of MIPS machine code …

[Brian Miller]

Excellent start!

This is really great stuff! I don't know it's development speed, but the Intel 4004 maxed out at 740KHz. The times, they are a-changin'!

[James Micallef]

Re: Excellent start!

Great stuff. If you think how long it took to get from the 4004 to even the Pentium let alone the i7, it will be a while until we get one of these on the desktop. However it does open the door to that future.

[Kunari]

Re: Excellent start!

There have been a bunch of lessons learned that could --perhaps-- be applied to CNT processors thus speeding up the development curve of CNT chips.

[Alan Bourke]

They really need a different three letter acronym ...

... because that one is just asking for japery.

[GreggS]

Re: They really need a different three letter acronym ...

ahh yes. Perhaps they could get a load of them to run in parallel and call them a bunch of CNT's

[Euripides Pants]

Re: They really need a different three letter acronym ...

Yeah, sooner or later someone will call the thing a CNT-pooter...

[Tom 7]

A way to go yet

In the late 1980's I had access to a 2.4Ghz (bipolar) process and, having found a design for a 600 transistor (fet) 16 bit CPU modified it and did some simulations of how it would perform - about 30X faster than a 20Mhz 386 that was available at the time.

I never made one cos there wasn't any suitable ram available! And I would've been shot!

If they ever get to mass produce carbon nanotube cpu's they may be of use but I have a feeling that we would be using things on a graphene level with maybe nanotube connections which might give more than an order of magnitude improvement over silicon which will be needed to make the AFM* investment worth while ...financially.

*Absolutely Fucking Massive

[John Smith 19]

Re: A way to go yet

"In the late 1980's I had access to a 2.4Ghz (bipolar) process "

GEC Marconi? I did not think MOSIS offered bipolar.

"I never made one cos there wasn't any suitable ram available! "

The Achilles Heel of any architecture.

[Caesarius]

Spin-off technology?

The fault tolerance, even fault correction, is a very important achievement. I wonder if this could be used to make e.g. silicon transistors smaller?

[Acme Fixer]

Re: Spin-off technology?

They may need to use this with silicon. When the dimensions get small enough to cause the defects to increase and the yields to go down, this may be a way of correcting the defects and rescuing the chips from the rejects bin. Just add a few spares here and there.

[Mr C]

tube transistor vs solid state

this could very well be the breakthrough similar in importance as the breakthrough from tube to solid state transistors.

It would lift us in a new era of semiconductors, which is a bad name but only for lack of a better one.

If they can make this work technology we cant think of now will become reality.

[Trevor 3]

Carbon Nano Tubes?

We are very lucky its Carbon Nano Tubes (CNTs) and they are not made of copper.

[Miek]

There's something about the sight of a chip wafer that makes me all gooey inside!

[MojoJojo]

Mistake?

"But the experimental nine-nanometre-size carbon nanotube gates are dwarfed by the 22nm gates on the latest generation of silicon chips"

Other news sources are reporting that the CNT gates are around 9 mircons in size - i.e. about 500 times bigger than current silicon ones. But apparently shrinking them isn't a big challenge.

[John Smith 19]

"Other news sources are reporting that the CNT gates are around 9 mircons in size "

Now that makes the 1KHz clock frequency much more understandable.

But the real takeaway is that someone has put all the pieces together to create a full chip mfg process flow, including the design techniques to cope with what appear to be material specific failure modes.

Thumbs up for that. IIRC CNT's have been slowly getting into some existing processes over the years but this is a full on dedicated CNT process, and that's a big move.

[diodesign]

Re: Mistake?

"the CNT gates are around 9 mircons in size"

You're right, oops: the gates in this chip are about 8um but the tech doesn't have to stop there. As I understand it, 9nm CNT gates are possible. I've tweaked the wording to reflect this.

C.

[kyza]

But..

...can it run BF4 @ 4k/120fps?

[PeterM42]

Excess CO2

Can they make these nanotubes from carbon dioxide? Then tell the bonkers scientists that this will soak up all the surplus CO2 that mankind is ALLEGEDLY pumping into the atmosphere and causing global warming, wait, COOLING, no wait again, WARMING, - whatever - that is actually being caused by a not totally stable raging nuclear furnace located 93 million miles offshore.

[Pascal Monett]

"without having to look for imperfections or even know where they were"

Far from me the idea of criticizing this breakthrough - anything that can potentially increase my computing power rates an automatic thumbs up as far as I'm concerned.

That said, I am a bit worried about sentences like the one in the title. Computers use electricity, and electricity can be a finicky beast - which is why we try to herd it as much as possible when we make use of it. So, reading that they are making computing parts without even knowing about defects is somewhat disturbing to me.

I mean, couldn't this be cause for unexplained behaviour in program results when the chip is used in production ?

Could someone enlighten me ?

[Colin 4]

Re: "without having to look for imperfections or even know where they were"

I suspect they mean designing a manufacturing process which is self-correcting and will automatically eliminate defects - like the bit where they run a current through the CNTs and the closed ones vaporize, maybe they can do something similar for misaligned CNTs.

So they just apply these processes, and they know all the dodgy CNTs will have been blown away.

[2cent]

Make a carbon tube listen

I suspect that possible answer is sonic frequency to stand the tube on their ends and lock them to position.

Regulating the vibrational patterns, if the connections can stand the stress, would allow for multiple path functions.

It's an interesting premiss.

[Anonymous Coward]

Re. Make a carbon tube listen

Might work, someone should try it.

I once made some silver conductive paint this way by "shredding" silver gilding foil with an ultrasound cleaner.

What about integrating barium titanate based transducer arrays under the chip, start depositing the CNTs and then using the transducers to do the moving around ?

-AC