Silicon Chip at End of Life?

Terry Olaes

Terry Olaes

I Used to be the [H] News Guy
Joined
Nov 27, 2006
Messages
4,646
Physicists predict that the conventional silicon chip has no longer than four years left before it hits the physical limitations of the material. Alternatives such as carbon nanotubes are being investigated but scientists have been unable to build precise circuit patterns with nanotubes, limiting the application of the technology.

Just as Gordon Moore predicted in 2005, physical limitations of the miniaturised electronic devices of today will eventually lead to silicon chips that are saturated with transistors and incapable of holding any more digital information.
Click to expand...
 
It's the End of the Line alright, Once 22nm is reached, They're be no more die shrinks until someone makes a breakthrough. :eek: The cpu makers must have some sort of plan I'd hope.
 
I'm sure Intel has this stuff in the works, and most likely IBM too. They plan 10 years ahead. There will be carbon nanotubes, or biological transistors, etc. Don't fret.
 
I agree with the last couple posts. There is always something around the corner. Hard drive manufacturers make new strides each generation, SSD technology is growing for instance, and processor technology is sure to find a new method. If its not multi-core as a solution it will be the nanotubes route or one of the many story leads we read about at [H].
 
Seems like about every 5 years or so we get another story like this. We currently have 45nm chips shipping. 10 years ago that was likely thought to be beyond the limits. The problem with predictions like these is that they assume "all else being equal" which is never the case. Obviously the laws of physics aren't going to change, but advances in materials science will ensure that we continue to have faster, smaller, more capable CPUs over time. And if the end truly is at-hand for Silicon, so be it, breaking out of an entrenched paradigm typically leads to even greater advances as scientists and engineers are forced to rethink how to do things.
 
Isnt there also research in Fiber Optic and Quantum processors?
 
Sovnade said:
I'm sure Intel has this stuff in the works, and most likely IBM too. They plan 10 years ahead. There will be carbon nanotubes, or biological transistors, etc. Don't fret.
Click to expand...

IBM, Sun, and several other big players has been working with switching light for well over 5 years now. They claimed it would get to market around 2020, but that was 3, 4 years ago and I haven't seen anything else on it since.
 
Oh yeah, light switching too...forgot about that one. You're right, it's been a while since I've seen any news on that front either. Regardless, at the moment I'd much rather see die-shrinks and less power consumption be the main goal for the next couple years. Games today are almost as complex as financially feasible, and modern, cheap hardware (8800gt, core series, dirt cheap ddr2) can already play it. Now if my 500w peak system only pulled 200w instead, and saved me $25 a month on my electric bill...that would be nice.
 
Process size is only one factor in a very complex cpu, which in turn is only one component in a very complex computer. While the end of process shrinks may certainly slow advancement it will by no means end it. There are many other avenues that can be pursued to continue the trend of faster - cheaper that has been the underpinning of the computer industry.

What about faster bus speeds?
Faster clock speeds?
More efficient production technologies (ie 450 mm wafer fabs), lowering silicon costs and allowing more cpu cores at the same price point.
Faster memory and more of it.
More efficient software design to improve performance of existing hardware.
etc. etc.

If this does come to pass it doesn't spell the end of computer advancement, just an augmentation of it.
 
My uncle told me that one of his friends that works in Germany is trying to figure out a way to create a conductive plastic, that was it eliminates heat from the equation. I'm pretty sure it was plastic, might have to ask again..
 
JethroXP said:
Seems like about every 5 years or so we get another story like this. We currently have 45nm chips shipping. 10 years ago that was likely thought to be beyond the limits. The problem with predictions like these is that they assume "all else being equal" which is never the case. Obviously the laws of physics aren't going to change, but advances in materials science will ensure that we continue to have faster, smaller, more capable CPUs over time. And if the end truly is at-hand for Silicon, so be it, breaking out of an entrenched paradigm typically leads to even greater advances as scientists and engineers are forced to rethink how to do things.
Click to expand...




It seems more like every year we get these dooms day stories.I dont give them much credence at all.
 
Lazy_Moron said:
Isnt there also research in Fiber Optic and Quantum processors?
Click to expand...

There is a LOT of work being done on Quantum computers....unfortunately for you and me, that work is being done in secret for the purposes of national security.

I am currently taking a class on Quantum computing, which focuses on the development of certain algorithms for certain very specific purposes, like Deutsch's Problem.

While I don't know much about Quantum computing, from what I do know, the possibility of a general purpose quantum processor seems to be a LONG WAYS off...i.e. I would feel lucky to see it in my lifetime.
 
Manny Calavera said:
It seems more like every year we get these dooms day stories.I dont give them much credence at all.
Click to expand...

Only problem is the current limitation put on silicon is down to 22nm, its beginning to cost too much to develop anything smaller and who knows how difficult it will be to re-create reduncy at that level or how much more money companys like TSMC are willing to put into creating a way to push it smaller. As it seems the advancement of chips rely's heavily on smaller processes when it gets to the point of not being able to get any lower it will probably stop or start a move to a new generation of materials.

I'm betting on a hybrid material to take place before we see carbon take its place, who knows what they have planned.
 
CMON!! give up! LOL.. we need something to ward of technological singularity... physical limitations is nature's way of saying.. ARE YOU CRAZY!! ;Þ
 
omg its the end of the world as we know it.

Jeez they can make a wafer 8 inches in diameter , what if it all was a single 22 nm device? Silly of course, but just because the density limit may be coming close to being reached with current methods does not mean you cannot just have a bigger die. It is not as efficient for the manufacturer (less dies per wafer) but for crying out loud, think how much circuitry at 22nm you could put on a die the size of the original Pentium.
 
At 22nm a 100 core CPU wouldn't be all that undoable. Parallelism will be the answer short term.

And I have to agree that the doom stories are always wrong. The 45nm process we have now, was little more than unimaginable Sci-fi 10 years ago. 10 years from now??? Who the heck knows. All we can say is that it will be exciting to see.

I would also agree that tackling some of the other ineffecient buses would be nice. How about a nice 1:1 ratio across all busses: CPU, FSB, Memory. All at 5ghz? Across a 16 core CPU? At 22nm?That would be quite nice.
 
Interesting, we've stuck with what we know for such a long time now that it's hard to imagine anything else, but the end of the silicon transister might be the push we need to invest in other technologies such as quantum computing. You can bet your bottom dollar that any advancement there is going to mean a substantial jump in processing power.
 
JethroXP said:
Seems like about every 5 years or so we get another story like this. We currently have 45nm chips shipping. 10 years ago that was likely thought to be beyond the limits. The problem with predictions like these is that they assume "all else being equal" which is never the case. Obviously the laws of physics aren't going to change, but advances in materials science will ensure that we continue to have faster, smaller, more capable CPUs over time. And if the end truly is at-hand for Silicon, so be it, breaking out of an entrenched paradigm typically leads to even greater advances as scientists and engineers are forced to rethink how to do things.
Click to expand...

Those have always been about manufacturing capabilities. The same materials have been used since the 1970's. When it starts becoming clear that manufacturers are reaching the end of their production capabilities, they try to find a solution that'll allow them to shrink even further (such as emersion lithography, SOI, strained silicon, copper interconnects, etc). But the active material has always bee silicon.

However, this is a physical limitation of the silicon material itself. Any less than 22nm (I think it's actually like 15nm) you start hitting problems where the atoms of the material itself interfere with the signal you're trying to push through. In fact, just to acheive 32nm manufacturers are going to have to use a new material as the gate layer just to prevent that sort of thing. Intel just jumped on using the new material early for their 45nm node, although it wasn't technically required for them manufacture at that size.
 
ryan_975 said:
Those have always been about manufacturing capabilities. The same materials have been used since the 1970's. When it starts becoming clear that manufacturers are reaching the end of their production capabilities, they try to find a solution that'll allow them to shrink even further (such as emersion lithography, SOI, strained silicon, copper interconnects, etc). But the active material has always bee silicon.

However, this is a physical limitation of the silicon material itself. Any less than 22nm (I think it's actually like 15nm) you start hitting problems where the atoms of the material itself interfere with the signal you're trying to push through. In fact, just to acheive 32nm manufacturers are going to have to use a new material as the gate layer just to prevent that sort of thing. Intel just jumped on using the new material early for their 45nm node, although it wasn't technically required for them manufacture at that size.
Click to expand...

Intel is using a new gate insulator material because the leakage characteristics of SiO2 are horrible for anything below 180nm. HfO2 is hi-K, so there is a much large barrier for the charge to jump, and the gates are metal so they can actually work with the HfO2.

Actual channel lengths at 45nm are about 26-27nm.. so that's 50 atoms across. When we start getting down to ~10 atoms, we'll start having real problems. ~16nm will be where we actually start seeing the end of silicon. The wildcard is the strain. We're using compressive strain on PFETs and tensile strain on NFETs. How big 10 atoms is depends on how well we can control strain.

For a reference on how fantastically cool strained silicon is, there is an equivalent pressure of 1000 MPa (1 Billion Pascal) on the gate of each transistor.
 
There will be something new. There just has to be. People have to think outside the box, aka organic computers and stuff like that. Hell, in the future a new computer might just be a pill you swallow with nano-machines that integrates into your body and display straight to the visual cortex. Come people, you seriously don't think we're gonna be stuck with silicon forever do you?
 
I am aware this is an ancient thread, but its mid 2011 and 22nm Ivy Bridge is just around the corner. Although Intel produced a 6 year roadmap, there's still no information in regards to silicon's successor.

2011: 32nm Sandy Bridge
2012: 22nm Ivy Bridge
2013: 22nm Haswell
2014: 16nm Broadwell (shrink of Haswell)
2015: 16nm Sky Lake (Sucessor to Haswell)
2016: 11nm Skymont (shrink of Sky Lake)

http://1.bp.blogspot.com/-Zact7IYIEAs/TfoPiofxwXI/AAAAAAAALuU/-MYA3wJQ2UQ/s1600/intel2017details.png
 
You must log in or register to reply here.
Back
Top