Intel: 'EUV Facts Don't Add Up' for 22 nm in 2011

[Idontcare]

Intel: 'EUV Facts Don't Add Up' for 22 nm in 2011
David Lammers, News Editor -- Semiconductor International, 4/22/2008 8:38:00 AM

Intel Corp. (Santa Clara, Calif.) has decided that extreme ultraviolet (EUV) lithography will not be production-worthy by 2011 when Intel plans to begin manufacturing 22 nm microprocessors, said Mark Bohr, director of process architecture and integration at Intel's logic technology development group (Hillsboro, Ore.).

?The facts don?t add up for EUV. Fact one is that Intel plans to do 22 nm in 2011. Fact two is that I don?t think anybody will claim that EUV will be ready for volume production that year. Maybe a year later, in 2012. I hope so, and Intel certainly will be pushing for that. However, we won?t delay our 22 nm technology to wait for it,? Bohr said.

http://www.semiconductor.net/a...A6553758.html?nid=3572

So EUV isn't necessary for 32nm or 22nm...provided you have immersion litho and some aggressive mask optimization protocols. Good news for just about everyone who was intending on going to 22nm (IBM, AMD, etc) but who likely weren't going to be capable of affording $180M for an EUV tool (IBM, AMD, etc).

Also two nice tidbits were dropped at the end of the story:

?We are on schedule to ramp 32 nm MPU products in the second half of 2009,? Bohr said. The 22 nm ramp will come in 2011, although challenges remain.

(OK, that one isn't much news, but confirms the schedule at least...Westmere likely coming in Q4/09)

And:

Asked whether Intel will be able to extend planar transistors to the 22 nm node, Bohr said, ?Whether it is planar or vertical, these are the kinds of questions we are asking ourselves. There are a couple of paths, including planar, that look like they can work at 22 nm.

Understand that planar transistors wouldn't even be an option to contemplate at 22nm node unless 32nm was guaranteed to be planar transistors. So that pretty much kills the earlier speculation (Nemesis) that 32nm will be anything more spectacular than a standard shrink (meaning no new materials or fancy 2D->3D changes) for the front-end...and quite possibly it will be extended to 22nm node as well.

Making the 16nm node the earliest timeframe for introducing 3D transistors as well as EUV. All continues to be good news for the competition as it continues to suggest the future technology roadmaps are not going to require Intel sized R&D budgets to make the next two nodes at least.

 

[Neverm1nd]

Originally posted by: Idontcare
...who likely weren't going to be capable of affording $180M for an EUV tool (IBM...

Ibms revenue 08Q1 was 24 billion, Intels 9.7
Ibms net income 08Q1 was 2.4 billion, Intels 1.4

 

[Nemesis 1]

Thanks for the link. I feel the need to remind you that Few if any thought High K and metal gates were on Intels 45nm. I did tho it was in intels long term guidlines at 45nm just as 3D gates was on the guidelines for 32nm.
Intel did a great job of keeping High K metal gates on 45nm secret. Until they announced at IDF. So I am sure they will keep 3D gates secret until spring IDF 09.


As for what Bohr said . I well bold the interesting part.

Asked whether Intel will be able to extend planar transistors to the 22 nm node, Bohr said, ?Whether it is planar or vertical, these are the kinds of questions we are asking ourselves. There are a couple of paths, including planar, that look like they can work at 22 nm.


Intel is pretty certain that IBM gate first tech on 32nm will be a finfet planar transistor.

Its anyones guess but I will stay with Intel 3D gates @ 32nm.

 

[bryanW1995]

Originally posted by: Neverm1nd

Originally posted by: Idontcare
...who likely weren't going to be capable of affording $180M for an EUV tool (IBM...

Ibms revenue 08Q1 was 24 billion, Intels 9.7
Ibms net income 08Q1 was 2.4 billion, Intels 1.4

yes, but ibm does a LOT more than sell cpus. what proportion of that revenue and NI is from their cpu/chipset business? IBM doesn't want to wager their company on going toe-to-toe with intel.

 

[Idontcare]

Originally posted by: bryanW1995

Originally posted by: Neverm1nd

Originally posted by: Idontcare
...who likely weren't going to be capable of affording $180M for an EUV tool (IBM...

Ibms revenue 08Q1 was 24 billion, Intels 9.7
Ibms net income 08Q1 was 2.4 billion, Intels 1.4

yes, but ibm does a LOT more than sell cpus. what proportion of that revenue and NI is from their cpu/chipset business? IBM doesn't want to wager their company on going toe-to-toe with intel.

IBM's semico business does not even rank them in the top 20...they have one (1) production fab.

http://www.isuppli.com/news/default.asp?id=8675

(I did not think I needed to post this as a qualifier to my assertions in the OP...but I guess I underestimated the audience a tad)

 

[Cogman]

Originally posted by: bryanW1995

Originally posted by: Neverm1nd

Originally posted by: Idontcare
...who likely weren't going to be capable of affording $180M for an EUV tool (IBM...

Ibms revenue 08Q1 was 24 billion, Intels 9.7
Ibms net income 08Q1 was 2.4 billion, Intels 1.4

yes, but ibm does a LOT more than sell cpus. what proportion of that revenue and NI is from their cpu/chipset business? IBM doesn't want to wager their company on going toe-to-toe with intel.

IBM could kick the trash out of Intel if they so desired, and I don't think they would have to bet their entire business. IBM has been partner to a fair amount of advances in CPU technology (even though they aren't a real competitor). Some of which gave AMD the advantage in the K8 era.

That being said, IBM doesn't have much competition in some of its nitch markets like specialized CPUs ect, so they haven't had anyone to fight against.

 

[degibson]

Originally posted by: bryanW1995

Originally posted by: Neverm1nd

Originally posted by: Idontcare
...who likely weren't going to be capable of affording $180M for an EUV tool (IBM...

Ibms revenue 08Q1 was 24 billion, Intels 9.7
Ibms net income 08Q1 was 2.4 billion, Intels 1.4

yes, but ibm does a LOT more than sell cpus. what proportion of that revenue and NI is from their cpu/chipset business? IBM doesn't want to wager their company on going toe-to-toe with intel.

IBM and Intel make very different computing products these days. Its not clear that IBMs best way to make money is to seek the latest and greatest process technology, because doing so isn't as cost-effective in their domain. Its cheaper to let Intel pioneer the field and then learn from their mistakes.

 

[Idontcare]

Originally posted by: degibson
IBM and Intel make very different computing products these days. Its not clear that IBMs best way to make money is to seek the latest and greatest process technology, because doing so isn't as cost-effective in their domain. Its cheaper to let Intel pioneer the field and then learn from their mistakes.

It's pretty much always been their business model from as far back as I experienced it...try and be first to research and somewhat develop a given process technology (CMP for example), scoop up all the early and easy IP and get your patents in order (IBM leads in annual patent filings) and then sit back and let the ROW (rest of the world) figure out how in the hell to get the stuff to work (yield well) in production environment while IBM collects the licensing revenues and bides their time before attempting to put it into their own production fab.

SiLK is about the only bone-headed thing I ever saw IBM stubbornly insist on making work in their production fab against the grain of the ROW. And when it became clear to IBM why the ROW did not pursue SiLK (90nm, Xilinx parts started failing early in the field) IBM quickly rushed to put blackdiamond into their Fishkill fab. Such recent and fresh memories of failure in customer sockets will keep IBM from doing anything that isn't proven acceptable for mainstream for a while. Too many executive bonus plans depend on it not happening again.

 

[zsdersw]

Originally posted by: Cogman
IBM could kick the trash out of Intel if they so desired, and I don't think they would have to bet their entire business.

With 1 production fab? Not a chance. The justification for investing in more fabs would be what?

 

[Idontcare]

Originally posted by: zsdersw

Originally posted by: Cogman
IBM could kick the trash out of Intel if they so desired, and I don't think they would have to bet their entire business.

With 1 production fab? Not a chance. The justification for investing in more fabs would be what?

IBM has been on a decidely obvious pathway of divesting itself of all the "capital intensive" manufacturing aspects of its prior business model.

Look at Microsofts PFO versus Intels PFO. There is something to be said about making your buck in the software and apps support business segments versus pushing capex into manufacturing chips (even when you have the advantage of being monopolies).

SUN and Qualcomm have shown that it doesn't require owning your own fabs to be good at what you do best. So long as IBM owns it's fab and the associated R&D expenses the corporate executives will have to continue to justify the existance of that business unit in light of the much more highly profitable and gross margined software and support business units.

The executives at Texas Instruments had to do this too...and eventually they threw in the towel after completing the 65nm node.

It's never a question of whether IBM can make money by continuing to re-invest in their fab...its purely a question of whether IBM can make even more money by investing those dollars into any other business segment.

When the answer has been painfully "yes" for too many business quarters then the shareholders (mutual funds managers) will get riled enough to press the point enough that executive management will have little to no choice but to "take action".

With TSMC getting ready to offer HK/MG for 45nm and 32nm SUN microsystems chips (they are inheriting the SUN account from TI since TI announced no further CMOS development after 65nm) this is opening the door for everyone else (IBM and AMD) to finally have access to a decently high performing (we hope) process technology at a foundry...and enables asset-smart opportunities to gain higher chances.

 

[zsdersw]

Nice post, but the relevance to my comment isn't clear.

 

[bryanW1995]

I think that he was replying to cogman. you, idc, and I appear to be operating in the real world while he is...not.

 

[Idontcare]

Originally posted by: zsdersw
Nice post, but the relevance to my comment isn't clear.

As bryanW1995 commented, my post was to further add to the dialogue you created in your post.

I quoted you more out of respect for the fact I was only adding the info to the thread because your post motivated me to do so.

My post was not directed towards you, or any specific poster for that matter, but rather just adding more text to the thread for would-be lurker readers.

 

[Lonyo]

Originally posted by: Idontcare

Originally posted by: zsdersw
Nice post, but the relevance to my comment isn't clear.

As bryanW1995 commented, my post was to further add to the dialogue you created in your post.

I quoted you more out of respect for the fact I was only adding the info to the thread because your post motivated me to do so.

My post was not directed towards you, or any specific poster for that matter, but rather just adding more text to the thread for would-be lurker readers.

Speaking of adding more, you got any links to the TI stuff?

 

[Idontcare]

Originally posted by: Lonyo
Speaking of adding more, you got any links to the TI stuff?

Sure...I worked there when it all went down (as in I was a process development engineer working on 45nm and 32nm nodes at the time this was announced). Here are some of the publicly stated stuff:

http://www.eetimes.com/news/se...ml?articleID=197000041

(note this is the original statement from Jan 2007...as such it states TI would continue to develop 45nm which turned out to not to be true after the course of the following 15 months)

http://www.washingtonpost.com/...0/AR2008022001322.html

(this one notes SUN signed TSMC for 45nm...which internally at TI was all that was keeping 45nm development going, once this news broke that ended TI's need for an internal 45nm node across the board)

 

[zsdersw]

Originally posted by: Idontcare
As bryanW1995 commented, my post was to further add to the dialogue you created in your post.

I quoted you more out of respect for the fact I was only adding the info to the thread because your post motivated me to do so.

My post was not directed towards you, or any specific poster for that matter, but rather just adding more text to the thread for would-be lurker readers.

Ah.. thanks

 

[Idontcare]

Cymer touts 100 watts average power in time for 2009 production ready EUV tools

Cymer said at this weeks Sematech EUV Source Workshop in New York that it has reached a milestone 25 watts of average extreme ultraviolet (EUV) power continuously for 1 1/2 hours duration, and is on track to achieve the 100 watts of average power needed for production-ready EUV lithography tools in 2009.

http://www.fabtech.org/content/view/6455/

2009 seems so close, can't believe Cymer is touting they will be ready to support production requirements of an EUV source in a matter of 12-18 months.

Pretty impressive progress. Let's hope it costs less than landing a man on mars when it comes to market.

 

[bryanW1995]

landing a man on the moon was chump change...mars could very well bankrupt our country... we're probably going to need an international push for that, which, admittedly, isn't necessarily a bad thing at all. It could be like contact...cue koom-bay-yah music...