News & Analysis
Comment
KB3001
This is akin to gambling in a casino as far as I am concerned. It might pay off ...
Kresearch
Again, it is an EUV marketing chant. If with breakthrough, arrogant like ASML ...
Globalfoundries looks to leapfrog fab rivals
Dylan McGrath
9/20/2012 9:01 AM EDT
SAN FRANCISCO—In an unexpected move, Globalfoundries Inc. said Thursday (Sept. 20) it plans to offer 14-nanometer process technology featuring FinFET three-dimensional transistors in 2014, just one year after the foundry's 20-nm process is scheduled to enter production.
The move could allow Globalfoundries (Milpitas, Calif.) to leapfrog foundry rivals and pose a challenge to Intel Corp.'s semiconductor process technology supremacy.
"Globalfoundries is getting extremely aggressive in their manufacturing roadmap," said Len Jelinek, director and chief analyst at IHS iSuppli. "They recognize that in order to be in a technology leadership position foundry, they have to be in a leading position across all semiconductor manufacturing, including Intel."
[Get a 10% discount on ARM TechCon 2012 conference passes by using promo code EDIT. Click here to learn about the show and register.]
Globalfoundries said its 14-nm XM process combines a 14-nm class FinFET and elements of the company's 20-nm low power process. Executives said the 14-nm FinFETs have a 48-nm fin pitch, the same as the company expects Intel Corp. to feature on its 14-nm tri-gate process. Other feature size measurements are also identical to what Intel is expected to offer at 14-nm, they said.
The acceleration of its process technology development roadmap will likely give Globalfoundries a clear technology lead over other dedicated foundry suppliers. Taiwan Semiconductor Manufacturing Co. (TSMC) and United Microelectronics Corp. (UMC) have indicated that they plan to integrate FinFETs in their 20-nm processes, which UMC is set to put in production in the second half of 2014 and TSMC likely some time after.
Executives said the new process would give customers the technology needed to compete with Intel, the undisputed leader in semiconductor process technology. Intel began production of 22-nm devices with FinFETS—which Intel calls tri-gates—earlier this year.
Mike Noonan, executive vice president of worldwide marketing and sales at Globalfoundries, said the company pulled in its roadmap specifically to "intercept" Intel at 14-nm. "The goal is to give our customers the power and performance to compete with Intel," Noonan said.
A conventional planar FET (left) and a FinFET. FinFETs offer greater power efficiency because they operate at a lower Vdd and have lower leakage.
Source: Globalfoundries
FinFETs are three-dimensional, double gate transistors that have been the subject of research by companies and universities for more than a decade. FinFETs consume less power than conventional transistors. Globalfoundries maintains that its 14-nm XM technology is expected to deliver a 40 to 60 percent improvement in battery life compared with today's two-dimensional transistors.
|
|
|
|
|
Navigate to related information
lee.t
9/20/2012 10:02 AM EDT
I am so confident that they will get it right the first time, and not have any yield issues, or nonsense with models, or anything else that could possibly go wrong.
;)
Sign in to Reply
askubel
9/20/2012 12:33 PM EDT
You're a little more optimistic than me. I think they're overconfident given that 14nm is the first node they'll use FinFETs at, along with the fact they haven't begun any kind of volume with 20nm (even 28nm is fresh out the door).
But we'll see. If Globalfoundries pulls this off they'll be in a strong position.
Sign in to Reply
Paul1960
9/20/2012 4:53 PM EDT
askubel, look up "irony" then read lee.t's comment again...
Sign in to Reply
askubel
9/20/2012 6:10 PM EDT
Note my careful use of comparatives in the first sentence :)
Sign in to Reply
Kresearch
9/20/2012 6:19 PM EDT
Hope it won't be marketing chant again. Go Globalfoundries.
Sign in to Reply
dylan.mcgrath
9/20/2012 6:38 PM EDT
I think you have to give GF credit for boldness. As the comments above allude to, they had their share of yield issues at 32-nm. The GF executives I spoke with said they wanted to wait until the process design kit was available to make the announcement. To G. Dan Hutcheson of VLSI Research, the fact that the process design kit is available is the key to the announcement.
Sign in to Reply
SKYHAWK
9/20/2012 8:42 PM EDT
this is to keep AMD from bolting to rival TSMC, so the CEO can keep his job awhile longer.
Sign in to Reply
m00nshine
9/20/2012 9:43 PM EDT
Reminds me of bold claim from TSMC about 450mm and ASML about EUV. Easy to make the bold claim, hard to actually do like that. I am a skeptic, but interested to watch this one.
Sign in to Reply
Kresearch
9/21/2012 12:18 AM EDT
EUV is dying. Triple patterning seems economically dead. 14nm will be very challenging without litho breakthrough. Bravo Foundry guys.
Sign in to Reply
motti2
9/21/2012 9:05 AM EDT
EUV little comprehended has found a breakthrough high average energy source ( ~100x greater than laser plasma EUV sources ) and you might end up eating your words in 2-3 years when the source integration is completed. There was an announcement by a U?Washington plasma physics prof about a Zpinch EUV plasma source ( discharge is magnetically confined by a current sheath ) that makes a simple longer pulse source of extraordinary net average power increase for EUV illumination.
I have little understanding of the technical issues relating to the source integration into the ASML EUV stepper, but the source itself is apparently a possibly compelling breakthrough.
Within weeks of disclosure of the source breakthrough by its professor inventor I think I remember that ASML issued an announcement that timeline for EUV "deployment" has been pulled in .... ( implying they will be using the source, but not mentioned at all )
I think there is a change that production viability prospects for EUV might have improved significantly, but this is as yet just decent speculation, based on ?coincidences of a technical nature ( from a process engineer )
Sign in to Reply
peter.clarke
9/21/2012 11:02 AM EDT
I think you are refering to Zplasma Inc.
http://www.eetimes.com/electronics-news/4376726/Nuclear-fusion-research-forms-plasma-for-EUV
Sign in to Reply
resistion
9/21/2012 11:50 AM EDT
Maybe you better get some more details on this. You don't want longer pulse, you want larger duty cycle and shorter pulse duration. And it turns out the resist dose window will not even benefit from a power boost.
Sign in to Reply
Kresearch
9/21/2012 5:54 PM EDT
Again, it is an EUV marketing chant. If with breakthrough, arrogant like ASML won't invite Intel,TSMC, Samsung to invest her recently. Intel even urged ASML to expedite ArF immersion development in 450mm lithograpgy. I guess we would stick on 20nm for a long while and 28nm or 20nm node would be a historically sweet spots for foundries.
Sign in to Reply
rick.merritt
9/21/2012 1:39 AM EDT
Time will tell if this is a bold leap forward or a foolish jump off a cliff.
Sign in to Reply
Huey
9/21/2012 7:52 AM EDT
That's what I think.. with leaning ---..
Sign in to Reply
Archeologist
9/21/2012 4:32 AM EDT
They are just renaming their 20nm process as 14nm. The challenge still remains to get yield with FinFETS.
When they do get yield they can say they are competitive. Without yield, it may as well be a mee too PhD thesis.
Sign in to Reply
BillSalmond
9/21/2012 7:10 AM EDT
Is this a joint development with IBM? If so, will other "fab club" members deploy this process(e.g. Samsung). IMHO Global do not have the resources or talent to do this on their own.
Sign in to Reply
SiliconAsia
9/21/2012 12:16 PM EDT
Let them claim what they want. The time is ticking....
Sign in to Reply
Roba66
9/21/2012 1:01 PM EDT
The backers of Globalfoundries have deep pockets but I think competing with Intel is an unhealthy obsession for them. I like profitable foundries that keep the mature processes going so people can make lots of mixed-mode chips.
Sign in to Reply
KB3001
9/23/2012 8:30 AM EDT
This is akin to gambling in a casino as far as I am concerned. It might pay off but it's still foolish.
Sign in to Reply