Intel willing to fab for competitors

[positivedoppler]

http://www.bloomberg.com/news/2013-...actories-for-other-chipmakers.html?cmpid=yhoo

"Otellini had said Intel’s plants were closed to competitors. Earlier this year, Krzanich said he would consider changing that stance, and yesterday he confirmed the company’s willingness to make chips for companies that are beating Intel in mobile phones."

What do you guys think this means? 14nm apples?

 

[BallaTheFeared]

Corporate espionage :ninja:

 

[frozentundra123456]

They sabotage all the chips so they self destruct in 2 months?

Seriously though, kind of surprising if it is on their smallest process node. Maybe it will be on say 22nm to utilize capacity after a new node is introduced? Seems especially strange since intels own 14 nm is delayed due to yield problems.

 

[ViRGE]

Seriously though, kind of surprising if it is on their smallest process node.

Not really. Intel has traditionally been described as a fab company with a chip design addiction. So they'll do anything on the fab side if the money is right. The "right" money being enough to deliver 65%+ margins, as they like to have.

 

[JDG1980]

Intel's 32nm process is very, very solid. I suspect that AMD would get better results on that fab than with GloFo even at 28nm. Unfortunately, that albatross of a contract that Hector Ruiz signed pretty much rules that out.

 

[tviceman]

Would be amazing if AMD and Nvidia could migrate GPU's over the Intel's latest and greatest fabs. It will probably never happen though.

 

[Vesku]

I hear Admiral Ackbar is available as the Intel spokesman for this new open fab initiative.

Seriously, though, any company big enough or innovative enough to be a competitor with Intel is going to have a lot of trepidations.

 

[ShadowVVL]

I don't think amd has the money
I could be wrong since I don't know much about how such business goes down but I don't think amd or nvidia would want their designs in the hands of their biggest competitor.

 

[ShintaiDK]

Would be amazing if AMD and Nvidia could migrate GPU's over the Intel's latest and greatest fabs. It will probably never happen though.

If it did, GPU prices would rise rather harshly.

 

[tviceman]

If it did, GPU prices would rise rather harshly.

I think die sizes could come in way smaller on 14nm vs. current sizes on 28nm and still have many times performance increases. I don't know how much per wafer Intel would charge, nor do I know how much die area would be saved to achieve equal performance on 14nm compared to 28nm, but it could offset costs. JHH has said in the past he would like to see Intel open up their fabs, so if there is an option Nvidia and AMD would be stupid to not at least look into it.

 

[SiliconWars]

TSMC's 20nm will be ready before Intels 14nm so that hardly matters.

 

[ShintaiDK]

I think die sizes could come in way smaller on 14nm vs. current sizes on 28nm and still have many times performance increases. I don't know how much per wafer Intel would charge, nor do I know how much die area would be saved to achieve equal performance on 14nm compared to 28nm, but it could offset costs. JHH has said in the past he would like to see Intel open up their fabs, so if there is an option Nvidia and AMD would be stupid to not at least look into it.

Plenty of others willing to pay much more than AMD or nVidia. And with limited supply...

 

[ShintaiDK]

TSMC's 20nm will be ready before Intels 14nm so that hardly matters.

So when do we see 20nm TSMC chips? GPU wise it looks like a whole year, if not more, before we see 20nm GPUs from AMD and NVidia.

 

[24601]

Plenty of others willing to pay much more than AMD or nVidia. And with limited supply...

The only ones that don't license most of their IP are Nvidia, AMD/ATi, Qualcomm and Apple.

Everyone else will have a hell of a time porting over their licensed designs tailored to TSMC fab processes.

This is shown in the fact that Intel is even fabbing externally with Sophia because of the significant importance of the TSMC validated design portion of their SoC.

 

[Maximilian]

http://www.bloomberg.com/news/2013-...actories-for-other-chipmakers.html?cmpid=yhoo

"Otellini had said Intel’s plants were closed to competitors. Earlier this year, Krzanich said he would consider changing that stance, and yesterday he confirmed the company’s willingness to make chips for companies that are beating Intel in mobile phones."

What do you guys think this means? 14nm apples?

So Intel can have a good long look at what's beating them and then develop something that beats it back. Then close fabs to competitors again :awe: Good plan!

 

[SiliconWars]

So when do we see 20nm TSMC chips?

Now, if you want it - http://electroiq.com/blog/2013/11/xilinx-ships-industrys-first-20nm-all-programmable-product/

GPU wise it looks like a whole year, if not more, before we see 20nm GPUs from AMD and NVidia.

This is AMD's and Nvidia's decision, and they'd presumably make the same decision on 14nm at Intel so no change.

 

[ShintaiDK]

Now, if you want it - http://electroiq.com/blog/2013/11/xilinx-ships-industrys-first-20nm-all-programmable-product/

This is AMD's and Nvidia's decision, and they'd presumably make the same decision on 14nm at Intel so no change.

So only extremely limited chips that get sold for 1000s of $? And not even listed on their site to buy?
http://www.xilinx.com/onlinestore/index.htm

So when do we get regular consumer chips on 20nm? AMD and nVidia is due to cost. Apple and Qualcomm will get 20nm before the 2.

 

[SiliconWars]

So only extremely limited chips that get sold for 1000s of $? And not even listed on their site to buy?
http://www.xilinx.com/onlinestore/index.htm

You asked when we'd see 20nm TSMC, I gave you the answer.

So when do we get regular consumer chips on 20nm? AMD and nVidia is due to cost. Apple and Qualcomm will get 20nm before the 2.

This would also, presumably, be the same on Intel's 14nm (if Qualcomm, Apple, AMD and Nvidia all decided to go build chips on it, Qualcomm and Apple would still be first due to cost).

 

[erunion]

TSMC's 20nm will be ready before Intels 14nm so that hardly matters.

So TSMC being a node behind and without finfets obsoletes Intel 14nm? Good try at thread crapping though.

 

[SiliconWars]

So TSMC being a node behind and without finfets obsoletes Intel 14nm? Good try at thread crapping though.

Thread crapping? The problem here is you're so fast to jump to the wrong conclusions that it blinds you to what is actually going on. I was obviously referring to tviceman's point about it being 28nm vs 14nm, nothing to do with "obsoleting" intel's 14nm or anything else.

 

[tweakboy]

dIs this news from fudz

 

[jdubs03]

We'll see 14nm Broadwell and Cherry Trail likely before Qualcomms A57 20nm chip, and around a similar time-frame as Apple's A8. Nvidia's Project Denver will likely be 1H2015. Samsung is less certain.

You keep trying to say Intel is going to be putting 14nm in production later than it really is, they said 1Q2014 not 2Q as you keep mentioning.

The yield problem is solved and improving, so even early 1Q is a possibility making at least an announcement at Computex possible.

 

[erunion]

Thread crapping? The problem here is you're so fast to jump to the wrong conclusions that it blinds you to what is actually going on. I was obviously referring to tviceman's point about it being 28nm vs 14nm, nothing to do with "obsoleting" intel's 14nm or anything else.

My apologies. I didn't see the connection to the previous post.

 

[SiliconWars]

You keep trying to say Intel is going to be putting 14nm in production later than it really is, they said 1Q2014 not 2Q as you keep mentioning.

The yield problem is solved and improving, so even early 1Q is a possibility making at least an announcement at Computex possible.

The yield problems are not solved. Improving yes, solved no.

The point I'm making is that this has been a feature for them so far on 14nm. Peaks and valleys one after the other. Ask yourself how likely it is that Intel is going to go ahead full-out production on a node that has shown these characteristics? It's not safe for them to go ahead on the hope that it's fixed - they need to know for sure.

So 14nm matches 22nm once in Q1? What if it drops again? No, they will make sure it's totally fixed.

So then (assuming they have 3-4 test batches of 22nm-level yield), at the end of Q1 they have to move the process from R&D to the actual fabs in OR and AZ. That takes time as well. After that's done it's generally 6 months before general availability.

Sure they can *maybe* paper launch something in Q2, but it'll be the mother and father of all paper launches. This is absolute best case by far - detractors will say 2015 and it's likely going to be in the middle, or late 2014 before we see any kind of decent 14nm availability.

 

[jpiniero]

Intel doesn't do paper launches though. Once the delay happened Broadwell was pretty much going to be only the high end QC models in 2014, so even if the yield sucks they should be able to deliver enough.