Samsung and GLOBALFOUNDRIES Forge Strategic Collaboration to Deliver 14nm FinFET

[ShintaiDK]

Why do you hate Samsung?

Eh?

 

[mrmt]

So the subpar foundry screwed 28nm, screwed 20nm and screwed finfet. And now they are back at square 0, licensing processes nodes from others, just like AMD used to do. 5 years and they are still unable to develop a node in a timely fashion.

I don't think the press is that idiot, so now we know who is in GLF payroll. This is a serious setback, and some of the press is selling this as a victory to GLF.

 

[ShintaiDK]

Samsung was/is in the same alliance as IBM. Samsung is just the "alliance leader" after IBM pulled out. But nothing changed besides that.

But it is interesting that GloFo essentially abandoned everything of their own. Lite foundry model?

 

[Phynaz]

The takeaway is that Samsung 14nm customers can port to GloFo with no effort, and use them as a true second source

Which is very interesting, Samsung makes it easier for Apple to walk away?

Glofo must have have written one very large check for this to occur.

 

[mrmt]

It looks like the big takeaway is-



http://www.bloomberg.com/news/2014-04-17/samsung-globalfoundries-team-up-in-challenge-to-tsmc.html

Seemless 14nm FinFET transition between fabs and production in 2H 2014.

The foundry competition with BILLIONS of chips produced annualy is nothing but good news. intel doesn't have near the volume to be able to compete with that.

Would you mind comparing rhe volumes, Mr. 100 billion?

 

[Bateluer]

So the subpar foundry screwed 28nm, screwed 20nm and screwed finfet.

How do you figure GloFlo screwed up 28nm? It was just late.

 

[NostaSeronx]

How do you figure GloFlo screwed up 28nm? It was just late.

It wasn't late, it was under performing when compared to TSMC. 28nm Bulk from GlobalFoundries has halfish the current drive of TSMC. While also being more expensive to boot.

28nm GlobalFoundries was faster to market but no one wanted to use such an expensive node. Yah, Globalfoundries beat TSMC to 28nm.

 

[piesquared]

Another great article here with a video announcement of the collaboration. I don't think this is a big problem for TSMC but the competition should drive development ahead faster giving all foundry customers access to nodes from cutting edge 14nm FinFETs, 16nm FinFET's and 20nm planar to older, and mature cheaper nodes. Win Win Win!

http://www.brightsideofnews.com/2014/04/17/samsung-and-globalfoundries-collaborate-on-14nm-finfet/

In terms of the capacity itself and the fabs capable of delivering this capacity, there will be a total of 4 fabs in 3 locations delivering this 14nm capacity to customers. Those come in the form of three Samsung fabs and one Globalfoundries fab. The locations of these fabs are actually in two countries, the US and Korea, even though Globalfoundries does have fabs in Singapore and Germany as well. However, those are not part of Globalfoundries current 14nm plans so they aren’t counted as part of the fab capacity for 14nm.

Looks like there will be significant quantity available. What a bombshell of a game changer!

 

[NostaSeronx]

Dresden and Singapore foundries from GlobalFoundries are going for 14nm FDSOI. This will be announced later this year. So, you have GlobalFoundries off-sourcing;

Samsung's 14nm FinFETs
and
Samsung's 14nm FDSOI.

 

[mrmt]

How do you figure GloFlo screwed up 28nm? It was just late.

What do you mean by "just late"? In the semiconductor industry time frame is everything. Take Bulldozer, it was a disaster when it was launched in 2011, but it would be a nice processor if AMD could get that level of performance in 2008. Globalfoundries launching 28nm in 2012 is something, and would have made it a competitor to TSMC. GLF launching this same process in 2014 is just meh, TSMC has milked away the high end customers of that market and now there are only second tier/low costs customers to tape out 28nm designs. So yes, they screwed up big time their 28nm launch, to the point that the first 28nm products only launched when all the bleeding edge customers were all gearing up for TSMC 20nm.

Btw, this is the nail in SOI's coffin. It's now clear that whatever IBM and STM had in their pipeline, it was not enough even for a subpar foundry like GLF, Samsung 14nm is better, period.

Btw2, I wonder what are the implications for IBM sale of its foundry.

 

[piesquared]

What do you mean by "just late"? In the semiconductor industry time frame is everything. Take Bulldozer, it was a disaster when it was launched in 2011, but it would be a nice processor if AMD could get that level of performance in 2008. Globalfoundries launching 28nm in 2012 is something, and would have made it a competitor to TSMC. GLF launching this same process in 2014 is just meh, TSMC has milked away the high end customers of that market and now there are only second tier/low costs customers to tape out 28nm designs. So yes, they screwed up big time their 28nm launch, to the point that the first 28nm products only launched when all the bleeding edge customers were all gearing up for TSMC 20nm.

Btw, this is the nail in SOI's coffin. It's now clear that whatever IBM and STM had in their pipeline, it was not enough even for a subpar foundry like GLF, Samsung 14nm is better, period.

Btw2, I wonder what are the implications for IBM sale of its foundry.

I guess that means intel is screwed since they are late with 14nm. Now they need to rely on architecture since the foundries have caught up. And, as we can see from Jaguar matching or beating Silvermont, 28nm planar vs 22nm FinFETs, they've got some problems.

 

[Bateluer]

It wasn't late, it was under performing when compared to TSMC. 28nm Bulk from GlobalFoundries has halfish the current drive of TSMC. While also being more expensive to boot.

28nm GlobalFoundries was faster to market but no one wanted to use such an expensive node. Yah, Globalfoundries beat TSMC to 28nm.

Eh? TMSC had working 28nm ARM parts on the market in 2012. What GloFlo 28nm parts were available then?

 

[mrmt]

I guess that means intel is screwed since they are late with 14nm. Now they need to rely on architecture since the foundries have caught up. And, as we can see from Jaguar matching or beating Silvermont, 28nm planar vs 22nm FinFETs, they've got some problems.

Jaguar doesn't match or beat silvermont at all. Btw, results are out, and guess what, they lost more market share.

 

[NostaSeronx]

What GloFlo 28nm parts were available then?

2010, 28nm-HP was out before 32nm-(SHP)SOI.

GlobalFoundries has been fabricating 28nm-HP Cortex-A9 chips for some unknown customer since then.

 

[Sweepr]

I guess that means intel is screwed since they are late with 14nm. Now they need to rely on architecture since the foundries have caught up.

Gotta love the wishful thinking. Cant wait to see AMD's latest and greatest 2015 28nm APUs up against Intel at 14nm & the ARM(y) @ 20nm/16nm FinFET.

 

[witeken]

I guess that means intel is screwed since they are late with 14nm. Now they need to rely on architecture since the foundries have caught up. And, as we can see from Jaguar matching or beating Silvermont, 28nm planar vs 22nm FinFETs, they've got some problems.

Intel is only 1Q late with 14nm, they're still 3 years ahead of TSMC, and 14nm will launch in Q4 (Cherry Trail) while 20nm will launch in Q1 (A57, the successor of the infamous, power hungry A15).

BTW, you know that Jaguar has the same idle power consumption as 1 Silvermont core during high workloads (0.8W)?

 

[mrmt]

GlobalFoundries has been fabricating 28nm-HP Cortex-A9 chips for some unknown customer since then.

Irrelevant. Whoever got GLF chips didn't order significant numbers to shake the market. And it seems that the results were unimpressive enough to not lure other customers to Globalfoundries arms. It's the same as not having received any orders.

 

[NTMBK]

Which is very interesting, Samsung makes it easier for Apple to walk away?

Glofo must have have written one very large check for this to occur.

I doubt Apple gave Samsung a choice. 'Give us a second source, or we go to TSMC/Intel.' Only way I can see it making sense to Samsung.

 

[NostaSeronx]

And it seems that the results were unimpressive enough to not lure other customers to Globalfoundries arms. It's the same as not having received any orders.

It was a lack of marketing really;
28nm-HP was capable of 100% parametric yields in 2010. So, whoever got those Cortex-A9 chips they were the most perfect chips ever made.

- No defects
- No variation
- No yield issues

 

[mrmt]

It was a lack of marketing really;
28nm-HP was capable of 100% parametric yields in 2010. So, whoever got those Cortex-A9 chips they were the most perfect chips ever made.

So GLF had a perfect functional node ready to be offered to customers BEFORE TSMC, but yet they couldn't get customers to TRY that node? Do you think we are idiots?

 

[NostaSeronx]

So GLF had a perfect functional node ready to be offered to customers BEFORE TSMC, but yet they couldn't get customers to TRY that node? Do you think we are idiots?

I think it was actually more or less;
GlobalFoundries node was SUPER EXPENSIVE.

While TSMC had a much cheaper and higher performing node even if it had defects, yield issues, variation. At a much lower cost.

 

[Abwx]

It wasn't late, it was under performing when compared to TSMC. 28nm Bulk from GlobalFoundries has halfish the current drive of TSMC. While also being more expensive to boot.

28nm GlobalFoundries was faster to market but no one wanted to use such an expensive node. Yah, Globalfoundries beat TSMC to 28nm.

Their 28nm HPP was somewhat late but it has actualy better perfs than TSMC s 28nm and judging by real numbers , not fud or hearsay, it s no worse than Intel s 22nm.

http://forums.anandtech.com/showpost.php?p=36265992&postcount=181

 

[mrmt]

I think it was actually more or less;
GlobalFoundries node was SUPER EXPENSIVE.

While TSMC had a much cheaper and higher performing node even if it had defects, yield issues, variation. At a much lower cost.

That is NOT a marketing failure, but an engineering failure. GLF could not sell what they had developed because it sucked, not because the sales team didn't do its job.

 

[Phynaz]

I think some people in this thread are treating GloFo as if they are still AMD.

 

[Fjodor2001]

I think some people in this thread are treating GloFo as if they are still AMD.

Could you please clarify that?

Also, I think what happens to AMD is in fact quite interesting when judging the future for AMD. Because what GloFo (and TSMC) has to offer process-tech-wise is also what AMD will be able to offer.

Now that GloFo will be adapting Samsung's 14 nm, it also means that AMD can release 14 nm chips at a time quite close to Intel (perhaps only ~1-1.5 years after).