Intel process roadmap...

[Qbah]

According to Xbitlabs, Intel will be using 4nm (four) fabrication process by 2022. Sounds incredible

As bold as it sounds though, someone needs to tell them the world's gonna end in 2012 anyway

 

[Lean L]

The roadmap ends at 2022 because the next year 2023 will be the end of the line according to moore's law. By then two things will happen

1) transistors will reach the size of a single atom therefore there won't be any progress to be had
2) since transistors won't get smaller, you can only increase the wafer size, this leads to longer paths for information to travel

 

[Forumpanda]

WTB superconducting 3D IC's

 

[imported_Shaq]

I don't think they will get that far. They thought they would reach 10Ghz too. After 20nm it will get a lot tougher.

 

[stipalgl]

Very interesting stuff. Although this next bit is purely speculation on my part, I suspect that at some point in time within the next ten years or so, Intel and AMD will possibly have to look at other alternatives for creating processors.

I'm talking about the end of the line for the traditional semiconductor circuit with transistors, etc. much in the same way several years ago, other technologies were replaced with more modern ones (ie. vacuum tubes giving way to better alternatives).

I was discussing the subject of computers with an individual with a biomedical engineering degree (I understand it's not the perfect background in this subject) and he was hypothesizing a computer processor built in a liquid medium as opposed to what we see today. Similar to the way our bodies work in transferring things through the bloodstream.

Extremely fascinating ideas and fantasies but we often have to dream and fantasize in order for those thoughts to manifest into reality.

 

[yh125d]

Seems kind of plausible. I mean 12 years ago we were using pentium IIs at what, 250nm and 300mHz?

 

[tommo123]

could we see optical computing (using photons i think), i read something a while back on using DNA for very specific things and bacteria doing calculations. gotta wonder where we'll be in 12 years. if anyone has the cash to seriously push boundaries (as well as have the motivation) it's intel.

 

[geokilla]

I thought this Moore's Law ended at 22nm.

Seriously, I'd like to see Intel slow down so AMD can catch up. If they keep on going like this and AMD fails to catch up, Intel will have a monopoly.

 

[taltamir]

Originally posted by: tommo123
could we see optical computing (using photons i think), i read something a while back on using DNA for very specific things and bacteria doing calculations. gotta wonder where we'll be in 12 years. if anyone has the cash to seriously push boundaries (as well as have the motivation) it's intel.

that doesn't make any sense. electricity propegates at C anyways.

 

[Juncar]

I laughed when I saw that. The road map looks more like something that the marketing guys came up with. Intel engineers probably face palmed. They already have troubles with MOSFETs at current size, at 4 nm I don't know if its practical at all with silicon based chips.

 

[Lean L]

Originally posted by: taltamir

Originally posted by: tommo123
could we see optical computing (using photons i think), i read something a while back on using DNA for very specific things and bacteria doing calculations. gotta wonder where we'll be in 12 years. if anyone has the cash to seriously push boundaries (as well as have the motivation) it's intel.

that doesn't make any sense. electricity propegates at C anyways.

it's close to .97C but yes... thus I rolled my eyes when I read about light CPUs a few years ago

 

[JackyP]

No, reason to roll eyes. Photonics and the use of light is a serious avanue of research, at least when it comes to interconnects.
http://en.wikipedia.org/wiki/S...cs#Physical_properties
http://techresearch.intel.com/...es/Tera-Scale/1419.htm

 

[taltamir]

thats optical INTERCONNECTS, not optical COMPUTING.
from the same link you posted http://blip.tv/file/963436/

And the articles posit that silicone laser based tech will be cooler and more power efficient and cheaper to produce than current optical implementation and maybe even current wire implementation.

The big advantage of optics is the lack of interference and EMR emissions. This means significant power savings... but right now the cost is high.

I am not saying that there is no room for optics in computers. But the way the layman presents and understands it reminds me of a clip where a samsung guy explains how organic light emitting diodes produce better pictures because they can produce more "natural" colors because they are "organic". (note: those are not organic as in grown in vegetables, all of them are completely artificial materials produced with a hydrocarbon skeleton; aka the guy is full of it). I also don't see 100% optical chips. but yes, we will probably get optical interconnects between portions of the PC.

 

[ShawnD1]

Originally posted by: taltamir
reminds me of a clip where a samsung guy explains how organic light emitting diodes produce better pictures because they can produce more "natural" colors because they are "organic"

Can't blame samsung for trying. People who drop out of high school won't understand why a new technology is great, so it's easier to just lie to them and make things up.

 

[TuxDave]

Originally posted by: Juncar
I laughed when I saw that. The road map looks more like something that the marketing guys came up with. Intel engineers probably face palmed. They already have troubles with MOSFETs at current size, at 4 nm I don't know if its practical at all with silicon based chips.

Looks like something that someone whipped up with an excel sheet.

45nm

= PREVIOUS * 1/SQRT(2)
= PREVIOUS * 1/SQRT(2)
= PREVIOUS * 1/SQRT(2)
= PREVIOUS * 1/SQRT(2)
= PREVIOUS * 1/SQRT(2)...

I wonder why they stopped. If they only went 8 more years they could've written 1nm transistors!!! wooo

 

[Idontcare]

Originally posted by: TuxDave

Originally posted by: Juncar
I laughed when I saw that. The road map looks more like something that the marketing guys came up with. Intel engineers probably face palmed. They already have troubles with MOSFETs at current size, at 4 nm I don't know if its practical at all with silicon based chips.

Looks like something that someone whipped up with an excel sheet.

45nm

= PREVIOUS * 1/SQRT(2)
= PREVIOUS * 1/SQRT(2)
= PREVIOUS * 1/SQRT(2)
= PREVIOUS * 1/SQRT(2)
= PREVIOUS * 1/SQRT(2)...

I wonder why they stopped. If they only went 8 more years they could've written 1nm transistors!!! wooo

They are labeling the process technology node, not giving physical dimensions of any components in the devices.

At 65nm node we had Lg's of 25nm already. 4nm node in that table is by no means intended to be a reference to 4nm xtors.

At any rate, just like EOT for gate oxides are referenced to a non-existing silicon dioxide thickness of equivalent electrical characteristics, I fully expect the node labels to continue their iterative cadence but the underlying physical characteristics of the features will take on an effective dimensionality or equivalence.

The physical dimensionality of devices fabricated on the 4nm node may have little to do that of the 6nm, but the relevant electrical figures of merit will be incrementally improved over that of the 6nm node such that the new node label is warranted and justified.

The concept has worked well enough for gate oxide and EOT, it'll probably serve an equivalent purpose in the world of node cadence post planar CMOS too.

 

[Veramocor]

From wikipedia, but seems about right "For comparison, the lattice constant, or distance between surface atoms, of unstrained silicon is 543 pm (0.543 nm). " That means at 4nm the distance would be a ridiculous 8 atoms.

 

[drizek]

8 atoms ought to be enough for anybody.

 

[TuxDave]

Originally posted by: Idontcare

I fully expect the node labels to continue their iterative cadence but the underlying physical characteristics of the features will take on an effective dimensionality or equivalence.

I'm kind of hoping they will not actually. (regarding node labels)

 

[ShawnD1]

http://www.wired.com/wiredscie...08/04/scientists-buil/

British researchers have unveiled the world?s smallest transistor, which measures one atom thick and ten atoms across.

The newly announced transistor is more than three times smaller than the 32 nanometer transistors at the cutting edge of silicon-based electronics.

 

[Idontcare]

Originally posted by: TuxDave

Originally posted by: Idontcare

I fully expect the node labels to continue their iterative cadence but the underlying physical characteristics of the features will take on an effective dimensionality or equivalence.

I'm kind of hoping they will not actually. (regarding node labels)

It is kinda unavoidable though, the physical limitations of shrinking to atomic dimensions is right around the corner and yet the node labels serve a very useful marketing role (as does clockspeed, hence the "PR number" marketing model that crops up repeatedly in history). The fact we did this with EOT to no one's objection is more evidence (to me anyway) that we are headed that way.

 

[njdevilsfan87]

Carbon nanotubes? (I guess Intel has assumed when the time is to come, something will be ready to replace Silicon)

 

[Lemon law]

We all know intel, I am looking forward to those 100 GHZ intel chips for $2.00, but if you want the 600GHZ model, its $2000. Lordi, its always hard to keep up with the Jone's.

 

[Idontcare]

Originally posted by: Lemon law
We all know intel, I am looking forward to those 100 GHZ intel chips for $2.00, but if you want the 600GHZ model, its $2000. Lordi, its always hard to keep up with the Jone's.

Oh come on :roll: that is just the stupidest analogy to Intel's pricing model I have ever read

We all know it will more likely be 597 GHz for $2, 600GHz for $500, and 603GHz for $1000.



No way in hell would Intel maintain a 6x performance or clockspeed delta top-to-bottom across their SKU's. That's just so 1989 and i486.

(I hope no one reads this post and doesn't realize I am being sarcastic with that opening sentence)

 

[OCGuy]

Originally posted by: geokilla
I thought this Moore's Law ended at 22nm.

Seriously, I'd like to see Intel slow down so AMD can catch up. If they keep on going like this and AMD fails to catch up, Intel will have a monopoly.

You want to see advancement halt so someone else can catch up?


Interesting read OP.