Slashdot Mirror
NASA Wires Chips With Nanotubes
carstene writes "SpaceDaily reports that NASA has come up with a way to wire microchips with nanotubes instead of copper interconnects. Aparently this could keep Moore's law a reality well into the next decade."
← Back to Stories (view on slashdot.org)
Just wondering - but how much would NASA have spent to find this out? I mean It's common to see companies like IBM come up with stuff that is cool like this (like the copper idea a few years back). It seems to me that Intel doesn't actually come up with too many new ideas? (I mean sure there chips become faster but not amazing new things).
I could be wrong. Has Intel done anything this cool? Surely they would spend more money on R&D for processors (I would assume NASA spends more on Space?)
any info about this would be much appreciated.
My blog [.net, rants, general IT]
Can they patent something created using your tax dollars?
Some Nanotubes are excellent conductors and some are poor conductors - depends on the tube type. So far it has proven difficult to grow only one kind of tube.
The way out may be a redundancy - several tubes doing the same function.
Maybe they can use them in vertical connections - for stacking chips up - one onto another, with nanotubes connecting the layers. But the overheating of such compact assemblies would be problem.
I doubt that we will ever figure out - and I suspect that even if we did figure out we couldn't do much about it
I've measured resistance of a nanotube of approx. 200 nm in length and about 5-10 nm in diameter to be a few hundred kilo-ohms (sorry, don't have exact numbers with me). This was for temperatures from room (300 K) down to about 2 K. We were looking at verifying some initial claims by groups claiming that nanotubes were superconducting. they aren't.
make world, not war
Ahem...
...
People saying "unless you are a gamer home computers are more than fast enough now for what we want (internet/email/minor word processing) " are forgetting that
1 - Starting Word 2024 will require 1.5 TeraFlops because every key you strike will require the calculation of two 8192 bytes key and the exchange of 1024 security tokens / sec, and we have to get ready to cope with that
2 - My old and faithfull Dual PIII 1Ghz, that was once considered the fastest rig on my block is now just a piece of interesting junk that still allows me to play Quake and encode divxs at the same time, and LOTS of you just dream about doing it for real
3 - it's not because i'm not a basic luser that immediatly jump categories and becomes a Power User. And if you think a softcore gamer or a hardcore Quaker is a "Power User", you never saw a real 16 CPU machine being "stability tested" for a round or ten of Quake @1024 fps, or the fastest Divx encode ever (11 minutes 8p)...
4 - "internet/email/minor word processing" can be achieved since 486 DX2 66 with no problem and little fuss... I mean my mail Server/Firewall/Ftp/ Webserver/PDC is a Pentium 133 and it serves the need of 10 ppl...So stop complaining when we allow you the use of a 2 Ghz computer just so you can play Freecell @ 25 fps 8p
Would be BOFH, hoping for Admin job...
It takes 40+ muscles to frown, but only four to extend your arm and bitchslap the motherfucker
I do not believe the rejection of magnetic field was ever experimentally observed. He quit there and said that he "retired" but was willing to send his experimental stuff to other people if they are willing to take over the research. No takers IIRC.
Well, it seems they are using multi-wall nanotubes
with rather large number of shells. Then you can
pass enough current to blow out all semiconducting
shells and get a metallic conductor. I don't
know if they use this trick but that's what IBM
people have done a while back.
The real trick is positioning these nanotubes
and contacting them. I wonder what they do to
assure good electrical contact. Typically your
contacts will be the first to blow out and the
thing to limit electronic mobility. Plus
encasing the nanotubes in silica sounds like a
bad idea because these suckers are really
sensitive to external perturbations and may not
conduct as well under external stress.
Faster, smaller, lighter computers are usefull for spacetravel. Just because they sendt a man to the moon with an onboard computer with less calculating power than a cheap pocket calculator and a weight of about 70 lbs (in addition to the 17.5 lbs DSKY) don't means that we should be satisfied with that sort of perfomance in the future.
BTW, more info on the Apollo guidance computer can be found at "One Giant Leap: The Apollo Guidance Computer" for those interested.
Everything in the world is controlled by a small, evil group to which, unfortunately, no one you know belongs.
I think technically they can, but I don't think they do.
I've seen plans for tiny scanning lasers (for docking alignment) on one of nasa's many websites, and loads of other stuff to boot. There's also a host of other reserach papers available online.
-- Waht? Tehr's a preveiw buottn?
couldn't any atom in the valence group do as well? (I'm remembering my old chart of the elements and we could have silicon nanotubes too.)
MSBPodcast.com The opinions expressed here are my own. If you don't like 'em... Think up your own stuff.
Yes, perhaps they promise less resistance than copper interconnect of the same size, but isn't a diameter of 100nm actually a bit large? Can nanotubes shrink, or is their diameter a chemical requirement? According to the International Technology Roadmap for Semiconductors, copper wiring pitch should now in 2003 already be 245nm. So with 50% spacing between those nanotubes, you're not even talking a 2x improvement in size over current interconnect. What if the things are too big to be used as interconnect for those 35nm gates we're supposed to see in 2007?
Who do you get to be an expert to tell you something's not obvious? The least insightful person you can find? -J Roberts
Right now, the majority of space on chip is taken up by verious caches. A significant proportion of that space is taken up by wiring. Having much smaller wiring should allow much larger caches. A system with 8Mb on-chip cache (and a well-designed asynchronous algorythm for filling it) would hardly ever wait for the front-side-bus at all.
Sig:Why copyright isn't a fundamental human right
Hyperthreading was built into the very first Pentium 4 (Willamette) processor, who's design started around 1994 (maybe 1995, I don't remember) under the codename P68. Regardless of when the design started, the first chip was released in what, 2000? That's well before Intel hired most of the Alpha team and got IP rights to Alpha technology.
Now, I'm not saying the Intel invented SMT (hyperthreading), but they didn't really just take it from Compaq either.
Most of the Intel inventions are either not disclosed (trade secret), or are modifications of existing technology to make it commercially feasible. Much of the process technology Intel adopts (or not adopts) is due to cost considerations, not just processor performance. Therefore, while IBM probably has technically better process technology, Intel has better yeilds (lower cost per processor).
Some "inventions" created by Intel include:
- the first microprocessor (4004)
- the first commercially used 2-level adaptive branch prediction (Pentium Pro) (invented with research done by Prof Yale Patt and his students)
- USB
- PCI
- AGP
- PCI Express (most of these buses were done by working groups headed by Intel)
- the first commercially used post-decode trace cache (Pentium 4)
- lot's of low power techniques with Pentium M
There are a lot more, but usually not public.
Dan