Globalfoundries Accelerates Roadmap: 14nm Chips in 2014

Status
Not open for further replies.
[citation][nom]jdwii[/nom]I don't think they can do this, Look at their 32nm die at release and the issues it was having.[/citation]

Their 32nm node wasn't great, but they've had a few deals and such with Samsung, so they might pull this off. Samsung is most definitely not a slouch in manufacturing as far as I'm aware.
 
G

Guest

Guest
Amazing that we were at 65nm 4 years ago and now we are going into the 14nm territory in two years. So how close to quantum computing are we now?
 

dragonsqrrl

Distinguished
Nov 19, 2009
1,280
0
19,290
If GF can pull this off, it would allow them to catch up to Intel in terms of process size. I have some serious doubts though, as would anyone who's been following the recent manufacturing trends at GF. I mean realistically, two major process shrinks in a row, year after year? It seems above and beyond the capabilities of even the most advanced manufacturing companies, much less GF.
 
[citation][nom]jupiter optimus maximus[/nom]Amazing that we were at 65nm 4 years ago and now we are going into the 14nm territory in two years. So how close to quantum computing are we now?[/citation]

We're probably at least a few decades off of quantum computing getting much relevance in consumer computing.
 

commandersozo

Distinguished
Jun 20, 2008
55
0
18,630
[citation][nom]jupiter optimus maximus[/nom]Amazing that we were at 65nm 4 years ago and now we are going into the 14nm territory in two years. So how close to quantum computing are we now?[/citation]
The jump to quantum computing isn't a matter of getting current design standards miniaturized, it's a whole new paradigm.
 
[citation][nom]dragonsqrrl[/nom]If GF can pull this off, it would allow them to catch up to Intel in terms of process size. I have some serious doubts though, as would anyone who's been following the recent manufacturing trends at GF. I mean realistically, two major process shrinks in a row, year after year? It seems above and beyond the capabilities of even the most advanced manufacturing companies, much less GF.[/citation]

I'm sure that Intel and Samsung could do it. So long as GF doesn't wait until after their 28nm and 20nm processes are finished to work on the 14nm process, it could be done on time. All that it would take is using an different tech team to work it out than the one that works on 20nm and have them working at about the same time. I don't claim to knwo if they'll succeed or not, but it is possible and it's also even practical.
 

tomfreak

Distinguished
May 18, 2011
1,334
0
19,280
[citation][nom]jupiter optimus maximus[/nom]Amazing that we were at 65nm 4 years ago and now we are going into the 14nm territory in two years. So how close to quantum computing are we now?[/citation]Intel/AMD has been knowned to make large die size CPU as large as 300-400mm2 for quite some time already. Fabs are moving to 450mm wafer too, this lower the cost further and allow more chips to fit per wafer or bigger die size.

if we actually hit the limit of silicon, you u likely to see Intel/AMD start making larger die CPU until it become too big to be profitable b4 they switch to something else other than silicon. Intel Ivy is @ 160mm2 @ 22nm, Haswell retain the 4 core, so we still got some room to grow die size to 300-400mm2. *Nehelem 1156 is 296mm2 tho.
 
[citation][nom]Tomfreak[/nom]Intel/AMD has been knowned to make large die size CPU as large as 300-400mm2 for quite some time already. Fabs are moving to 450mm wafer too, this lower the cost further and allow more chips to fit per wafer or bigger die size.if we actually hit the limit of silicon, you u likely to see Intel/AMD start making larger die CPU until it become too big to be profitable b4 they switch to something else other than silicon. Intel Ivy is @ 160mm2 @ 22nm, Haswell retain the 4 core, so we still got some room to grow die size to 300-400mm2. *Nehelem 1156 is 296mm2 tho.[/citation]

It gets more difficult and more difficult to make such large CPU dies with smaller and smaller processes, especially without dropping frequencies at least a little. Die shrinks seem to shrink the size by much more than they shrink the power consumption at a given frequency, so it'd take serious CPU design improvements to counteract this.
 
G

Guest

Guest
How about they get 28nm working right first. Also, not all processes are made equal. Just because one company is at 14 doesn't mean their process is as good as another at 14. Solve the issues of yield first. Solve the issues that they will have going from gate first to gate last (Intel AND TSMC already do this). Then solve the FinFETissues as we know that is not easy based on Intel's difficulties. Then... then... we can talk about 14, but all that has to be solved before they can go to 22/20nm.

http://www.bit-tech.net/news/hardware/2011/01/20/ibm-and-globalfoundries-go-gate-last-for-20/1

GF is still WAY behind.
 
[citation][nom]TheTruthIs[/nom]How about they get 28nm working right first. Also, not all processes are made equal. Just because one company is at 14 doesn't mean their process is as good as another at 14. Solve the issues of yield first. Solve the issues that they will have going from gate first to gate last (Intel AND TSMC already do this). Then solve the FinFETissues as we know that is not easy based on Intel's difficulties. Then... then... we can talk about 14, but all that has to be solved before they can go to 22/20nm. http://www.bit-tech.net/news/hardw [...] t-for-20/1GF is still WAY behind.[/citation]

Oh no, they're way behind because they decided to change from gate-first to gate-last processes. Also, what difficulties is Intel having with their FinFET tech?

Furthermore, problems with one process doesn't mean that they'll have problems with much newer processes even if they're a smaller process.
 

bigdog44

Honorable
Apr 6, 2012
167
0
10,680
Don't get too excited. The 3D implies tri-gate, but it's planar and not the same design as Intels', which also implies lower transistor density and lower structural strenght of the circuits.
 

abitoms

Distinguished
Apr 15, 2010
81
0
18,630
Good to hear it is coming in 2014. but when in 2014? Q1, Q4 or something in between?

And ramping up to 14nm means actual retail 14nm chips will come out in 2015 right?
 
G

Guest

Guest
You know changing from gate first to gate last is not as easy as it sounds right? Intel had several issues moving to FinFET. Only one product line is even using FinFET currently (no Atom, no Itanium, etc.). None of this is simple. The fact that they have struggled going from 32 to 28nm doesn't bode well for these far more intrusive changes. I have my sincere doubts about 2014 as should anyone based on their track record. Yes, blazorthorn, they are way behind. No other way to articulate that. Just because they state something in an article doesn't mean it will come true.
 

TeraMedia

Distinguished
Jan 26, 2006
904
1
18,990
Who'd have thought that in the end the CPU competitor for Intel wouldn't be AMD, but rather the fab it spun off? If ARM and Samsung are throwing cash in the game to help R&D costs, then GF might actually have a chance at competing against Intel's fab R&D group. That in turn would help level the playing field for chip designers so that Intel's CPU design team doesn't start out 20-50% better than AMD or ARM before they've even drawn a transistor.

I wish them luck.
 

ojas

Distinguished
Feb 25, 2011
2,924
0
20,810
[citation][nom]dragonsqrrl[/nom]If GF can pull this off, it would allow them to catch up to Intel in terms of process size. I have some serious doubts though, as would anyone who's been following the recent manufacturing trends at GF. I mean realistically, two major process shrinks in a row, year after year? It seems above and beyond the capabilities of even the most advanced manufacturing companies, much less GF.[/citation]
Yeah i guess that's why Intel waits at least two years before the next shrink...
 
[citation][nom]ojas[/nom]Yeah i guess that's why Intel waits at least two years before the next shrink...[/citation]

Intel waits two years because it is convenient, not because they have to wait. Why have a die shrink every year when they can do just fine with a die shrink every two years or so? If it was difficult and they needed more time, then they'd just get separate teams working on each die shrink simultaneously and they could churn out a die shrink as often as they want to until nothing more can be had from die shrinking without some other technology changes.
 

hector2

Distinguished
Mar 18, 2011
61
0
18,630
[citation][nom]blazorthon[/nom]Intel waits two years because it is convenient, not because they have to wait. [/citation]

Not true. It take a lot of work to get the next node out of the lab and ready for production. 22nm actually was delayed a few months beyond 2 years

 
[citation][nom]hector2[/nom]Not true. It take a lot of work to get the next node out of the lab and ready for production. 22nm actually was delayed a few months beyond 2 years[/citation]

They could have simply started work on it earlier and that wouldn't have been an issue.
 

TeraMedia

Distinguished
Jan 26, 2006
904
1
18,990
@blazorthon: The reason Intel "waits" is because at each step, they identify a bunch of new problems that they need to address to make that step a success. Those same problems usually also apply to the subsequent steps too (plus the additional problems that those yet-smaller steps face). So if you have team 22 working on 22nm R&D, much of the team might get stuck in a holding pattern while a small research group including one or two members solve some tricky problem. But if you also have team 14 working in parallel, then you have all of teams 22 and 14 waiting for that same small research group. Repeat this throughout the R&D cycle for 22, and you end up wasting a lot of team 14's time... and, they still have to face the issues that they weren't experiencing at 22 nm but crop up at the smaller scale. You've probably heard the saying, "Nine women can't deliver a baby in one month."
 
Status
Not open for further replies.