Future Fab report focused on ITRS roadmap

[Idontcare]

http://www.future-fab.com/content/PDF/FF36_Jan_11.pdf

Figured I'd post this up perchance there were two or three fellow forum members who have interest in this sort of thing

I'm still reading through it (more like "perusing it" at the moment) but the good stuff, for me, seems to start on page 72. (I'm biased though, having worked with both Joel and Jeff IRL)

Definitely get the impression that 20nm and above is viewed as "in the bag", but everything below 16nm is very much still up in the air. Which is pretty much "situation normal" for this industry.

Process tech to make nodes N+3 work are always a guessing game, have been for decades and yet we always pull through

Personally I still find it amazing that Intel is the only company to be shipping HKMG enabled CMOS...I never thought the gap in the timeline would be THIS lengthy.

 

[mnewsham]

-pulls out another mountain dew- i'll report back in a few hours. if i don't i was swallowed whole by the large mass of text.

 

[WhoBeDaPlaya]

All I know as a designer is - "Quit f#(*(#@*$ shrinking things! I want my nice analog voltage headroom "

 

[jimhsu]

The thing about 16nm is that it's not a process tech that is the problem; it is the fundamental laws of physics (rather, quantum mechanics). Don't know too much about the field, but I do remember that the probability of electron tunneling across an infinite square potential is proportional to the inverse cube of distance - so going smaller makes tunneling FAR more likely = bad. All I know is that going below 16nm will require innovations on the order of things like Giant magnetoresistance heads for hard drives; in other words, we'll probably be stuck there for some time due to quantum tunneling concerns, but probably will make a huge leap once someone figures it out. 3D integrated circuits? Nanoelectronics? I don't know.

 

[Hacp]

The thing about 16nm is that it's not a process tech that is the problem; it is the fundamental laws of physics (rather, quantum mechanics). Don't know too much about the field, but I do remember that the probability of electron tunneling across an infinite square potential is proportional to the inverse cube of distance - so going smaller makes tunneling FAR more likely = bad. All I know is that going below 16nm will require innovations on the order of things like Giant magnetoresistance heads for hard drives; in other words, we'll probably be stuck there for some time due to quantum tunneling concerns, but probably will make a huge leap once someone figures it out. 3D integrated circuits? Nanoelectronics? I don't know.

If the potential is infinite, then shouldn't the probability of tunneling be zero?

 

[Idontcare]

If the potential is infinite, then shouldn't the probability of tunneling be zero?

LOL, and yes. P.I.B. solutions are not interesting for infinite square well boundary conditions

Finite potential well is the more interesting opportunity.

 

[Wizlem]

If the potential is infinite, then shouldn't the probability of tunneling be zero?

I think the well is a space between two infinite pillars of finite width. Any who, getting somewhere outside of any well has atleast non-zero probabiliy in quantum mechanics due to conditions of the wave function(if the wave function is zero in an uncountable number of points it has to be zero everywhere which means the particle doesn't exist in the first place).

As for the probability of crossing such a barrier, I thought it was an exponential decrease with distance.

 

[khon]

As for the probability of crossing such a barrier, I thought it was an exponential decrease with distance.

It is, which is why decreasing barrier width is becoming such a problem.

 

[nyker96]

I think you need to put in title: pre-req: a PhD in EE.

 

[cbn]

If problems at 16nm and 11nm do occur, I just have to wonder what Intel is planning.

At what point do other options start looking more economically feasible?

On one hand they have this little company ARM (and its partners) trying as hard as possible to take away market share. Microsoft has bailed on them with respect to the x86 alliance. The MS server chief was fired one day after CES. Google Android (running ARM) is making fast progress into larger screen devices.

One the other hand Intel has developed Meego (which actually runs ARM in addition to x86) and Intel AppUp as alternative ecosystem for their processors. But how long will this buffer Intel's x86 processor business when it looks the situation is beginning to go downhill for x86. Sure Sandy Bridge is a great processor (but lets remember that design was taped out back in 2009). What is Intel designing right now?

Of course, I am not saying Intel will fail. They are too large and diversified to let potentially diminishing sales in x86 bring them down. However, at the same time I am skeptical they wouldn't be planning uarch alternatives at this point? (eg, to beat the ARM competitor through design rather than process technology)