Intel plans transistor scaling down to 4nm :o

[josiahsuarez]

http://www.hardware.info/en-US/news/ymicl5qVwpuacJY/Intel_shows_roadmap_of_up_to_4nm_in_2022/

http://xs942.xs.to/xs942/09346/intel_tech_2022_01474.jpg

The roadmap features a light-blue seperation, to which Intel responded that the tick-tock strategy will be under fire after 2012. Though transistors of 16 and even 11 nm are theoretically possible at this point, the 8 nm process will require another technological breakthrough. It's still unclear how transistors at this size are to be realised. The blue part of the roadmap is to be perceived as an extrapolation of the results of past years.

[CaptNKILL]

Four? :o

[Muppet]

That would be some feat.

[Ninja Prime]

I don't expect them to get below 11nm myself, maybe even 16nm.

[josiahsuarez]

Pat Gelsinger made some comments a year ago where he seemed confident about the prospects for future scaling.

http://www.crn.com/hardware/208801780;jsessionid=P4KZFDOWRURQHQE1GHRSKH4ATMY32 JVN
Speaking about Intel co-founder Gordon Moore's eponymous "law" regarding the expected doubling of transistors per integrated circuit every two years, Gelsinger noted that there was a time when he and his Intel colleagues wondered if they'd ever be able to scale chips below 100 nanometers.

"But we did do that, and today we see a clear way to get to under 10 nanometers. With Moore's Law we always have about 10 years of visibility into the future, so beyond 10 nanometers, we're not sure how we'll do it," he said.

that makes sense. 10 years is a long time and predictions tend to get less accurate as they go farther off into the future. let's get to B and C before worrying about D and E :)

[AthlonXP1800]

Wow very exciting future technology outlook, 2nm will be out in 2024 and the last of nanometer 1nm in 2026.

So what after nanometer? Picometer! :D

Picometer is a trillionth of a meter, we will see the chips with picometer or pm in 2028. (nana2)

[Revs]

Nanotubes FTW!

[zer0]

new nanosuit FTL

[josiahsuarez]

interesting bit of trivia

http://www.wolframalpha.com/input/?i=atom+size+in+nanometers
Input interpretation:
convert length scale of atom to nanometers

Result:
0.1 nm (nanometers)

Additional conversions:
~~ 100 pm (picometers)
~~ 0x10^-10 meters
~~ 4x10^-9 inches
~~ 1 Ao (Angstrom)

[coldpower27]

Hmm so they plan to get down to 11nm at least... below even 10nm would be a feat.

[pakotlar]

Hmm so they plan to get down to 11nm at least... below even 10nm would be a feat.

What's interesting to note is that coincidentally Xbitlabs reported yesterday about technology that would allow this kind of lithography.

http://www.xbitlabs.com/news/other/display/20090819032439_Nuclear_Fusion_Research_Set_to_Enab le_Nanolitography.html

Here is a more technical study: http://www.engr.colostate.edu/ece/faculty/rocca/pdf/journals/ECEjjr00092.pdf

Basically current passed through lithium emitting 13.5nm spectrum plasma, allowing for sub 13nm features; the light spectrum employed is more than an order of magnitude smaller than that used today.

[josiahsuarez]

when Pat said "we see a clear way to get to under 10 nanometers" he's probably talking about a half-step beyond 11nm. and he might just be talking about SRAM rather than microprocessors. still, it's good to hear Intel isn't giving up on scaling without a fight :)

if someone can mass produce @ 8nm when everyone else is stuck on 11nm they will have a big advantage and certainly plenty of demand. so the financial impetus is definitely there. assuming it's physically possible of course. I would hesitate to say it's impossible though until someone has run up against that wall and says from experience "ok, we tried, it just can't be done." like Pat says <100nm was "impossible" in the past, now <11nm is "impossible." who can say about tomorrow.

[pakotlar]

when Pat said "we see a clear way to get to under 10 nanometers" he's probably talking about a half-step beyond 11nm. and he might just be talking about SRAM rather than microprocessors. still, it's good to hear Intel isn't giving up on scaling without a fight :)

if someone can mass produce @ 8nm when everyone else is stuck on 11nm they will have a big advantage and certainly plenty of demand. so the financial impetus is definitely there. assuming it's physically possible of course. I would hesitate to say it's impossible though until someone has run up against that wall and says from experience "ok, we tried, it just can't be done." like Pat says <100nm was "impossible" in the past, now <11nm is "impossible." who can say about tomorrow.

No, if they can scale SRAM, logical gates come next. I don't think there has ever been an SRAM shrink test, without having logic following. It'd also make no sense; if the lithography is there to produce 4nm SRAM, the tech is not for away from producing 4nm logic.

[Intel17]

This neatness is exactly why I want to become an engineer...so that maybe one day I can actually work on this stuff :D

[josiahsuarez]

No, if they can scale SRAM, logical gates come next. I don't think there has ever been an SRAM shrink test, without having logic following.

I'm not saying that SRAM will scale farther than logic. I'm just not aware of the full context of Pat's statement.

[john19055]

4nm ,what is that about 10 atoms.I don't see how they can make anything that small.

[josiahsuarez]

atoms vary in size. 4nm is equal to ~125 helium atoms, but only ~18 caesium atoms.

http://en.wikipedia.org/wiki/Atom#Size