What is intel and the other fabs current roadmap for 10nm?

[Roland00Address]

So I heard on some websites that Intel may be delaying by a few months specific fabs getting tool delievery neccessary for that fab to transition to 10nm. The prolbem is this stuff is rumors and it is hard to sort bad rumor from useful info. Because of this I am calling all people from this forums who are following the technical merrits but more important the expected roadmaps for Intel 10nm as well as the other fabs such as TMSC and Samsung.

What I am trying to get are these stages in development and roughly which quarter we expect Intel or the others to hit that milestone.

Questions I am asking

1) Actual fab tool delivery in the fabs that are going to make the chips so we can get. This is important for it is roughly when you get the first batch of chips from real life like conditions instead of things developed in more "theroetical" and "lab like conditions." Effectively the Alpha "benchmark"

2) Ramp production of the chips effectively switching the foundry from in development/fine tunning stage to real chips

3) Shipping chips to OEMs and Consumers. Can I buy a 10nm chip in late 2016 or do I have to wait till 2017?

4) Which fabs of each company are expected to be releasing X chips in this year. For example to my understanding intel is going to be using Fab 28 for 10nm, but what other fabs. Another example is Fab 42 was supposed to be making 14nm chips but intel decided they had enough 14nm capacity and then stopped ordering 14nm tools for 42 prefering to get that fab ready for lower geometries than 14nm.


I am just trying to make sense where Intel but also its competitors truly are on the foundry level. 14nm was a disaster on the actual shipping date, how bad is 10nm? Intel was promising in Sept 2013 that 10nm will be in shipping products during the year 2015, there is no way this is going to happen. Are we going to get any real info prior to IDF in August?

Thank you everybody.

PS I do know that Samsung and TMSC planned 10nm will have lesser characteristics likely than Intel 10nm for you no longer can compare the companies naming schemes and marketing, regardless I am just trying to make a general shape of what is really happening right now in the murky waters.

 

[VirtualLarry]

Questions I am asking

1) Actual fab tool delivery in the fabs that are going to make the chips so we can get. This is important for it is roughly when you get the first batch of chips from real life like conditions instead of things developed in more "theroetical" and "lab like conditions." Effectively the Alpha "benchmark"

2) Ramp production of the chips effectively switching the foundry from in development/fine tunning stage to real chips

Intel, as I understand it, uses a fab production methodology called "copy exact". They get fab processes fine-tuned in their development fabs, and then copy that process exactly to all of the production fabs that they are using for that node. As I understand it, 10nm is either well under way at Intel, or basically finished, as far as far processes go. So I'm pretty certain that their developmental fabs already received their tools. I have no idea about their 10nm production fabs, I guess that's part of what you're asking.

 

[jpiniero]

Seems very likely at this point that Intel will release in 2017, and could be mid 2017. The don't-call-it-layoffs suggest they want to milk 14 nm moreso than any problems with 10 nm but we will see.

 

[Systems analyst]

ARM/TSMC have given a date of 4Q2015 for 'Risk Production' of 10nm with;
45% power reduction at same speed.
Density increase of 2.2 times.
See TechCon 2014:
http://ir.arm.com/phoenix.zhtml?c=197211&p=irol-presentations.

See also:
http://www.arm.com/about/arm-and-ts...processors-on-10finfet-process-technology.php

ARM helps the FABs to customize ARM designs to the production process and node of the FAB concerned; POP IP is one of the products concerned; see www.arm.com.

There is some general discussion at:
http://www.dailytech.com/TSMC+Hypes...ggles+to+Hit+Volume+at+16+nm/article37298.htm

Time will tell.

 

[Roland00Address]

Intel, as I understand it, uses a fab production methodology called "copy exact". They get fab processes fine-tuned in their development fabs, and then copy that process exactly to all of the production fabs that they are using for that node. As I understand it, 10nm is either well under way at Intel, or basically finished, as far as far processes go. So I'm pretty certain that their developmental fabs already received their tools. I have no idea about their 10nm production fabs, I guess that's part of what you're asking.

I understand the principle of copy exact, but Intel pushed back its 2nd line factory with Intel based 42 by several months. So does that mean there is a delay in 10nm in general, or intel just does not think there will be market demand, or they were overestimating when they had 10nm ready?

Anway who is the original fab they are basing 10nm off of. Is it D1x2 in oregon or is another fab? D1x2 is supposed to come on in 2015 sometime did it come on, or is it still in development.

As you can see I have questions and thus me trying to find the answers via asking for help

 

[ShintaiDK]

It all happens in oregon. And D1XM1 and DX1XM2 will be the first to start production. The 2 fabs are also 450mm wafer ready.

(Oregon also contains D1D fab).

I am sure Intel is already running 10nm, but working on yields. Intel requies a large yield output before startign actual mass production unlike TSMC/GloFo. 10nm in late 2016 or early 2017.

Older map, but more info:

 

[Enigmoid]

Expect delays. Expect delays all around.

 

[witeken]

Lol, this EETimes article from 2009 literally confirms III-V for 10nn. At least that that's the plan. So the question is if they succeeded.I'd say yes.

http://www.eetimes.com/document.asp?doc_id=1172569

 

[witeken]

Meanwhile at the foundries, costs are stagnating.

http://www.eetimes.com/author.asp?section_id=36&doc_id=1326864&

 

[ShintaiDK]

Its going to be hard to justify 10nm designs that doesnt pull 1.5 billion revenue home. Ouch.