How small can they in reality go, 40nm is small

[john19055]

Don't they reach a point where you can not go any smaller without haveing major leakage between the plates.I mean 55nm is pretty small right now.They did'nt gain a Lot going from a 65nm to a 55nm IMO,besides about 30watts of less power usage.

[Ninja Prime]

Theoretically, the limit is 15nm. 32nm is certainly possible, because Intel has already created samples at that size. But who knows for sure beyond 32nm, major changes are going to need to happen to get to 25nm, and even greater to reach 15nm.

Edit: Based on this: http://news.zdnet.com/2100-9584_22-133066.html

[Bman212121]

Interesting. I assume 15nm is probably where there are no longer enough atoms to build a complete transistor. Makes you wonder once they get on 32nm if they will slow down scaling to smaller steps like 25, 20, 18, 16, 15nm. That would already get them another 5 years down the road.

EDIT: Well actually it sounds more like just leakage would be the main issue. They can still make a smaller transistor.

Here is an article from 2001.


SANTA CLARA, Calif. -- Intel Corp. here today announced it has broke its own record for the world's smallest transistor, claiming the development of a 15-nanometer device that will be used to make microprocessors and other chips by the end of this decade.

Intel's 15-nm transistor (gate-length)--which has been devised in the laboratory--is a CMOS-based, 0.8-Volt device, said to handle switching speeds of 0.38-ps--or 2.63 trillion switches per second. Previously, Intel also claimed the unofficial world's record last June, with the development of a 20-nm transistor capable of 0.75-ps switching speeds--or 1.53 trillion switches per second (see June 10 story ).


More importantly, Intel's 15-nm device is expected to become a key element in the development of high-speed chips, based on the company's so-called P1268 process technology. On Intel's roadmap, the company's P1268 process is a 30-nm (0.03-micron) technology that will make use of 300-mm wafers.


Intel is expected to develop chips based on this process by 2009. By then, the company could have high-speed processors running at speeds of 20-GHz or faster, according to analysts.


The development of the 15-nm transistor also proves that bulk silicon will continue to be a viable technology in semiconductor manufacturing for the foreseeable future, said Gerald Marcyk, director of Intel's Components Research Group.


"If you look at Moore's Law, we are trying to shrink the transistor 30% every two years," Marcyk said. "This technology will take us out at least until the end of the decade," he said in an interview with SBN.


The Intel manager was referring to the popular axiom in the semiconductor business--attributed to Intel co-founder Gordon E. Moore--which says the number of transistors in integrated circuits doubles every 18 months due to device shrinks and other chip-processing technologies.


Before the end of this decade--or sooner--Intel and other chip makers will face some major challenges to develop ICs with bulk silicon, especially in terms of power consumption, Marcyk said.


The challenge is to make devices with low "standby currents and standby power," he said. "Developing smaller and faster devices is not a problem. The challenge is to make devices smaller, faster, and with lower power," he said.


Intel faces some other issues as well. The company will continue to embrace silicon until the end of the decade. But it's unclear if silicon will remain a viable technology beyond the next decade. "This is as far as I can rationalize it," the Intel manager said.



http://www.eetimes.com/news/semi/showArticle.jhtml?articleID=10810046

Pretty sure neither the 15nm prediction or the 20Ghz prediction came true. :)

A couple of other interesting reads.


Researchers have built the world's smallest transistor - one atom thick and 10 atoms wide - out of a material that could one day replace silicon.

Companies like Intel have a roadmap to reduce the size of circuits on the silicon wafer, down to about 10 nanometres - 10,000 times smaller than the width of a single human hair.

Many researchers believe that producing circuits smaller than 10 nanometres in silicon will be too difficult because they start to leak electricity at that size.

That current silicon roadmap is expected to end in 2020, making the race to find alternative materials potentially very lucrative.

http://news.bbc.co.uk/2/hi/technology/7352464.stm

This has already happened I think, as the article is from 1999. It still has an interesting analogy in it.

The vertical transistor may solve this problem by using the thickness of a precisely-controlled layer of material, rather than light, to set the gate size.

"Suppose you have a can of paint and a big paintbrush, and you are asked to paint the thinnest possible line," Hergenrother said. "If you just tried to paint the line freehand, that would be similar to the light approach.

"However, if you paint a flat surface, cut it vertically and look at it on edge, you will see a line that's as thin as the layer of paint. A similar principle is used in our transistor to produce the smallest gates ever made with the control that industry requires."


http://news.bbc.co.uk/1/hi/sci/tech/528109.stm

[josiahsuarez]

1nm is 3-5 atoms. that's not enough to transmit the amount of electrons needed for an integrated circuit to function. but the ITRS goes down to 11nm. see here:

http://en.wikipedia.org/wiki/11_nanometer
The 11 nanometer (11 nm) node is the technology node following 16 nm node. The exact naming of this technology node comes from the International Technology Roadmap for Semiconductors (ITRS). According to the 2007 edition of this roadmap, by the year 2022, the half-pitch (i.e., half the distance between identical features in an array) for a DRAM should be 11 nm, although Intel's "Architecture and Silicon Cadence Model" places it closer to the year 2015. Intel's Pat Gelsinger claims that Intel sees a 'clear way' towards the 11 nm node.

so we have about 10 years of transistor scaling left before we hit the "red brick wall" (the point at which further scaling is impossible). 11nm may or at not be an ultimate limit, but there is definitely an ultimate limit out there somewhere:

http://books.nap.edu/openbook.php?record_id=10582&page=32
A new technology will be needed, because, beyond the numerous technological and economic difficulties arising with sub-10-nanometer minimum dimensions, MOSFET integrated circuit (IC) technologies cannot shrink indefinitely. Ultimately, the discrete atomic nature of matter limits the shrinkage.

at that point the way forward is currently unclear. one possibility is using multiple layers of semiconductors, essentially taking the computer chip into the third dimension. new problems with cooling arise with that technology though (a heatsink would cool the top layer, but what about the others?). IBM is doing some interesting research on this subject:

http://www-03.ibm.com/press/us/en/pressrelease/24385.wss
IBM (NYSE: IBM) Researchers, in collaboration with the Fraunhofer Institute in Berlin, demonstrated a prototype that integrates the cooling system into the 3-D chips by piping water directly between each layer in the stack.

These so-called 3-D chip stacks – which take chips and memory devices that traditionally sit side-by-side on a silicon wafer and stacks them together on top of one another -- presents one of the most promising approaches to enhancing chip performance beyond its predicted limits.

http://xs537.xs.to/xs537/09126/chip_final_holeclosed_72704.jpg

[Ninja Prime]

Here is the article I got my info from: http://news.zdnet.com/2100-9584_22-133066.html

[Bomb30]

1nm is 3-5 atoms. that's not enough to transmit the amount of electrons needed for an integrated circuit to function. but the ITRS goes down to 11nm. see here:

http://en.wikipedia.org/wiki/11_nanometer


so we have about 10 years of transistor scaling left before we hit the "red brick wall" (the point at which further scaling is impossible). 11nm may or at not be an ultimate limit, but there is definitely an ultimate limit out there somewhere:

http://books.nap.edu/openbook.php?record_id=10582&page=32


at that point the way forward is currently unclear. one possibility is using multiple layers of semiconductors, essentially taking the computer chip into the third dimension. new problems with cooling arise with that technology though (a heatsink would cool the top layer, but what about the others?). IBM is doing some interesting research on this subject:

http://www-03.ibm.com/press/us/en/pressrelease/24385.wss

anyone else find it creepy how that diagram looks like the chip from terminator2? skynet here we come!!

[john19055]

That is incredable to think that something like 60 to 80 atoms could do anything being that small.I was just wondering because 55nm is very tiny and to think they can even go smaller ,down to 15nm just boggles the mind or at least mine.

[CaptNKILL]

That is incredable to think that something like 60 to 80 atoms could do anything being that small.I was just wondering because 55nm is very tiny and to think they can even go smaller ,down to 15nm just boggles the mind or at least mine.

No kidding. The fact that we can mass produce things that mindblowingly complex with near perfect accuracy is also... well... mind blowing. :p

Seriously. Its amazing that there are 781 million transistors inside a Core i7 CPU but its absolutely incredible that humans designed machines that can pump out thousands of them with little to no defects, all the way down to the nanometer scale.

[mullet]

No kidding. The fact that we can mass produce things that mindblowingly complex with near perfect accuracy is also... well... mind blowing. :p

Seriously. Its amazing that there are 781 million transistors inside a Core i7 CPU but its absolutely incredible that humans designed machines that can pump out thousands of them with little to no defects, all the way down to the nanometer scale.

This is the main reason why Skynet will come true in our life times. :(

[josiahsuarez]

heck, even 1 micron (1,000 nanometers) is mind blowingly small. that's 0.0001 centimeters! and Intel has been making chips at that scale since the 386 in the 80s D:

[AirRaid]

32nm and 22nm shouldn't be too difficult. I think, or from what I understand, 16nm will be a real challenge, and beyond that, who knows.

[XDanger]

Why cant they just make em really big?

[ViN86]

Theoretically, the limit is 15nm. 32nm is certainly possible, because Intel has already created samples at that size. But who knows for sure beyond 32nm, major changes are going to need to happen to get to 25nm, and even greater to reach 15nm.

Where did you get this number from? What is the restriction due to? Atoms are on the order of Angstroms (~.1 nm). So, a few hundred atoms makes sense at 15 nm. I was just wondering what this number was actually caused by.

Also, as we get smaller, EM interference from surrounding transistors continues to rise and new materials with higher dielectric constants are needed. They are in development currently, but they could place a damper on Moore's Law if the materials engineering doesn't keep up.

[CaptNKILL]

Why cant they just make em really big?

Larger pieces of silicon are more expensive and require more power which creates more heat. This is why they can cram more transistors into a smaller chips while actually improving efficiency over larger chips that are less complex.

At least, that's my understanding of it.

[Ninja Prime]

Where did you get this number from? What is the restriction due to? Atoms are on the order of Angstroms (~.1 nm). So, a few hundred atoms makes sense at 15 nm. I was just wondering what this number was actually caused by.

Also, as we get smaller, EM interference from surrounding transistors continues to rise and new materials with higher dielectric constants are needed. They are in development currently, but they could place a damper on Moore's Law if the materials engineering doesn't keep up.

I linked the article a few posts after, I probably should go back and edit it in to the actual post. Here it is:

http://news.zdnet.com/2100-9584_22-133066.html

Basically, quantum tunneling at that point where gates are only 5nm becomes too big of a problem unless you increase power drastically, at which point you cause more heat and defeat the purpose of making a faster chip, because you have to downclock to keep from melting the chip. If this was somehow solved(unlikely, IMO), its possible to go to 1.5nm gate size, which would be a transistor about 6nm, at which point you can no longer actually extract a charge from such a small well.

[Bman212121]

Also, if you missed one of my articles from above, it is definitely worth a look.

http://news.bbc.co.uk/2/hi/technology/7352464.stm

Using a different material it is possible to build a transistor that is only a few nm wide. It is about 5 years newer than the ZDNet article, so a lot of things could have changed since then. That other article I linked to from 1999 predicted 15nm and 20GHZ processors by 2009. (10 years)


EDIT: Actually this sums it up pretty well.

The transistor, essentially an on/off switch, has been made using graphene, a two-dimensional material first discovered only four years ago.