Submitted by Scientists at IBM say they have figured out how to produce smaller and more powerful microchips than previously thought possible. It is hoped IBM’s announcement at San Jose on Monday will mean the creation of miniscule microprocessors which will save the IT manufacturing sector billions of dollars. The breakthrough revolves around the distance between the circuit-lines chip makers must ‘draw’ onto the surface of a computer processor. IBM scientists declared they can now draw lines on silicon much closer together than ever before.
They expect to be able to decrease to 32 sm? Assuming that’s 0.032 nanometer, that seems a pretty bold assumption if 29.9 nm is such a huge breakthrough. It appears the actual limit is approaching.
Quote: They expect to be able to decrease to 32 sm?
No, it’s obviously a typo in the article, they’re talking about nanometer. There’s no such thing as “sm”.
So they figured out how to fix the problem with crosstalk and electromagnetic interference. They can only go so much farther before the laws of physics kick in and Moore’s law dies.
Soon enough, circut traces on motherboards will be gold strands a few molecules thick. Once that is reached, then what?
By that time you will have most stuff using hispeed serial connects.
Then you hope it becomes inexpensive enough to layer chips creating cubes of chips. Cube = 100 chips ?1000 ?
Seeing how IBM and AMD co-operate, and if the Intel way of “Not invented here” is true, then it looks like Intel will have a harder time in the future to keep their large market share.
Then again, who knows what Intel will come up with…
That’s not really true. Intel is beating AMD to 65nm production. AMD seems to skip a nm generation or two between plant refurbs being a little less aggressive then Intel when it comes to chip tech. Will they get it? Sure, but probably not first. So far they’ve done just fine anyway
Intel has been quicker with smaller nm sizes, but the advantages of this might be used in part for larger nm sizes, and then AMD would get more out of the same size as opposed to an Intel core. (which it seems like they already do)
Maybe it’s time to give up transistors:
http://www.wired.com/news/technology/0,70190-0.html?tw=rss.index
WOW! That is fscking amazing! I hope that NAND CPUs come to pass soon.:)
While it sounds cool, using only NAND (and it says NOR) operations means that most logic will take more logic gates to complete. That’s why nobody’s made a commercial NAND-only processor yet
Also, this tec is already behind in the size area being 110nm while AMD is at 9xnm and Intel will have 65nm soon.
That paper tell me very little about how such magnetic domains could function as logic gates without me having to do more google research. I would almost class this as an April 1st paper but not so sure. The one good reference it gives to the IBM pdf looks reasonable but thats about using 10nm domain dots for storing data on disk drive surfaces & not logic so there is no way to figure this paper for now. I’d wait for more credible references before concluding anything from this.
For now Silicon based transistors have plenty more life in one form or another, and magnetic domains for storage has same.
can achieve the impossible. It’s a well known truth. What gets investments always gets improved. _Anything_ that gets investments, funds, research will improve. Positively. Scientists will find a way to overcome all obstacles – they always do. One way or the other. Remember going to the moon? That was considered impossible too. If they set the goal of keeping Moore’s law intact and do whatever it takes to do it, then it’s going to happen. My take is that the scale of integration is going to get even less than 10nm and that this is going to be the _easy_ part. The interesting part will be when new non-(exclusively-)-silicon based technologies make their way to mainstream computing. We already have samples of this stuff, but when they hit mainstream this kind of stuff is going to change our lives. What we’ll be dealing with until 2010 can be considered as predictable `future-history’. Boring stuff.
When I started out in semiconductors in 79, it was obvious to everyone you couldn’t draw anything smaller than the wavelength of light if you use light to do the patterning. At that time 3u was quite a few wavelengths and the coming 1u was expected to be the point when ebeam would take over then followed by xrays. many paprs were published then on ebeam as the succesor to optical patterning. Well we all got it wrong, 1st patterns got smaller than 1 wavelength, then partial, so then deeper uv light or shorter wavelength with special optical materials.
I can barely understand how this works now but an understanding of fourier analysys helps. Basically what is drawn is not what is desired, instead you draw anti features that will be distorted by the processing in such a way that the end result will look like the desired result. A bit like writing backwards with a highly distorted mirror to complicate things.
Still its amazing to those in the industry too. Once we have to use xrays then Moore’s law will get expensive.