Samsung shows off 14nm FinFET wafers and systems

[NTMBK]

http://semiaccurate.com/2014/12/03/samsung-shows-off-14nm-finfet-wafers-systems/

Test wafers are back then, yay. No mention of yields of course.

 

[NostaSeronx]

No mention of yields of course.

Worse than 28nm in 2011.

 

[krumme]

If its already 2011 28nm yields its pretty good. No need for fdsoi tech then.

 

[NostaSeronx]

If its already 2011 28nm yields its pretty good. No need for fdsoi tech then.

I'm talking about GlobalFoundries 28nm yields. Sorry, for not specifying, I was hoping people will get I was referencing another CPA foundry.

"If it ain't SOI it will never be on time." - GlobalFoundries 2012

 

[Arachnotronic]

http://semiaccurate.com/2014/12/03/samsung-shows-off-14nm-finfet-wafers-systems/

Test wafers are back then, yay. No mention of yields of course.

The yields are on track. Duh.

 

[witeken]

The yields are on track. Duh.

What does that -- oh, you're being sarcastic.

 

[krumme]

Suddenly yield was all important...
A popular word here at AT forums since november 2014. Yield, yield, yield all over.
Introduced the day Samsung was thought to be Intel. (We talk about the all important EUV 10nm midnode logic - Happening h2 2016)
Or also known as The Hans De Vries giant whoopass day.

Somebody should be cured. But still someone knows the yield from looking at the pictures. Where can i buy those contact lenses?


For all we know this is set to be in the 2015 h2 apple a9. If yield is okey in a year then fine for ss or they will have to print some more. They aparently have started. Good enough for iphone it seems.

 

[Idontcare]

http://semiaccurate.com/2014/12/03/samsung-shows-off-14nm-finfet-wafers-systems/

Test wafers are back then, yay. No mention of yields of course.

Reminds me of the early days of GloFo when they trotted out 28nm and 20nm test wafers at every opportunity and made it seem like they were about to school TSMC on the art of being a foundry.

Marketing needs to learn some new tricks, this one we've already seen before.

 

[krumme]

I'm talking about GlobalFoundries 28nm yields. Sorry, for not specifying, I was hoping people will get I was referencing another CPA foundry.

"If it ain't SOI it will never be on time." - GlobalFoundries 2012

Gf is competing with ss here. I dont know how reliable info is.
As for ss they tell whatever fits larger strategy so who knows what today brings, but for sure the ramp is very steep so yield must be very low even if they are on track.
But what do i really care. I get my 14nm finfet h2 2015 in my note 5. If yield is low samsung shareholders pay. Fine for me.

 

[Arachnotronic]

Suddenly yield was all important...
A popular word here at AT forums since november 2014. Yield, yield, yield all over.
Introduced the day Samsung was thought to be Intel. (We talk about the all important EUV 10nm midnode logic - Happening h2 2016)
Or also known as The Hans De Vries giant whoopass day.

Somebody should be cured. But still someone knows the yield from looking at the pictures. Where can i buy those contact lenses?


For all we know this is set to be in the 2015 h2 apple a9. If yield is okey in a year then fine for ss or they will have to print some more. They aparently have started. Good enough for iphone it seems.

"Mass production" is utterly worthless without good yields; that's why people care about it.

 

[NostaSeronx]

Reminds me of the early days of GloFo when they trotted out 28nm and 20nm test wafers at every opportunity and made it seem like they were about to school TSMC on the art of being a foundry.

The 14nm test wafer from the image looks sort of like the 28nm test wafer from GlobalFoundries. It must be an ARM thing to have everything look the same from a wafer standpoint.

---
I rather look out for 28XHP/20XHP and 28XLP/20XLP, than look at 14LPP/14LPe.

 

[oobydoobydoo]

Rumors 12.2 inch "ipad Air Plus" with A9 in june, but the Note 5 should be out before then, right? I am guessing the first 14nm product will be the apple watch, but the Note 5 is a good bet too.

http://www.techtimes.com/articles/2...umors-project-q2-2015-launch-with-a9-chip.htm


Edit: I totally forgot about Core M. Intel already has 14nm.


Speaking of, a family member works for a certain company that will be incorporating intels 14nm process into it's new products. According to him, Intel's 14nm is in fact a much better process than Planar 20nm. The issue they are having is any node under 20nm increasing leakage a lot and it becomes very difficult to get increases in performance along with density. He says that while in the past you would see 30% increase in performance and 30% power reduction along with 2x density, now you will only see density improve. So its not looking good for any of the foundries unless they move to something other than silicon. This family member is very high up in this company and he knows what is going on, and I trust him. Hopefully all of you do too

 

[krumme]

Reminds me of the early days of GloFo when they trotted out 28nm and 20nm test wafers at every opportunity and made it seem like they were about to school TSMC on the art of being a foundry.

Marketing needs to learn some new tricks, this one we've already seen before.

The difference is GF cartoons was funny (at first). Samsung is kind of a dangerous brutal animal - with 1000 times more heft as backup.
The situation on the business side couldnt be more apart.

 

[Arachnotronic]

The difference is GF cartoons was funny (at first). Samsung is kind of a dangerous brutal animal - with 1000 times more heft as backup.
The situation on the business side couldnt be more apart.

Samsung is powerful, but they seem just as guilty of the typical foundry hype as everybody else.

Bottom line is TSMC is raking in the vast majority of foundry profits and it's not by accident. Samsung is putting on a good show, and they may well win a big chunk of the A9 volume (perhaps by offering low priced wafers because their foundry business is in such a dire position right now relative to TSMC), but TSMC is still going to be making most of the money in foundry, IMO.

 

[Arachnotronic]

Rumors 12.2 inch "ipad Air Plus" with A9 in june, but the Note 5 should be out before then, right? I am guessing the first 14nm product will be the apple watch, but the Note 5 is a good bet too.

http://www.techtimes.com/articles/2...umors-project-q2-2015-launch-with-a9-chip.htm

Seriously doubt that the Apple Watch, which is probably based on some Cortex M-something-or-other is going to be the first product on an expensive, bleeding edge 14nm node.

Note 5 or a Galaxy Tab S refresh would be the best bet if not Galaxy S6.

 

[oobydoobydoo]

Seriously doubt that the Apple Watch, which is probably based on some Cortex M-something-or-other is going to be the first product on an expensive, bleeding edge 14nm node.

Note 5 or a Galaxy Tab S refresh would be the best bet if not Galaxy S6.

I think they are using a bespoke chip designed by Apple. I doubt they would throw a cortex A9 into something they charge $4999 for, but perhaps. I do think 100% apple will use THE BEST process because the Apple watch is designed to last for years, and they will need all the processing power they can get for it.

I am sure that Samsung will use 14nm for the S6, so that may be first, but I don't doubt that apple will also use 14nm on the watch.

 

[krumme]

Samsung is powerful, but they seem just as guilty of the typical foundry hype as everybody else.

Bottom line is TSMC is raking in the vast majority of foundry profits and it's not by accident. Samsung is putting on a good show, and they may well win a big chunk of the A9 volume (perhaps by offering low priced wafers because their foundry business is in such a dire position right now relative to TSMC), but TSMC is still going to be making most of the money in foundry, IMO.

I dont know what ss is doing. And why. Beats me - seems crazy. If anyone can explain the strategic sense of it i would like to hear it?

But they are going all the way and they pay for it big time as evident from their exploding r&d budget. And now we start to see indications of results. Surely there must be good reasons for them investing so much in the foundry business.

 

[Arachnotronic]

I think they are using a bespoke chip designed by Apple. I doubt they would throw a cortex A9 into something they charge $4999 for, but perhaps. I do think 100% apple will use THE BEST process because the Apple watch is designed to last for years, and they will need all the processing power they can get for it.

I am sure that Samsung will use 14nm for the S6, so that may be first, but I don't doubt that apple will also use 14nm on the watch.

Fair enough argument; should be interesting to see how this all pans out.

 

[Sweepr]

Galaxy S6 should be out by early April 2015, I think that's way too soon for 14nm FF. They only have two 20nm planar SoCs out there (Exynos 5430 and 5433) and those are available in specific models of the Galaxy Alpha and Galaxy Note 4 (most units are based on Snapdragon). I could be wrong but I think their first 14nm Soc will power either 2015 Galaxy Tab S line (they need a very powerful chip to compete with A8X/A9/A9X) or the Galaxy Note 5 @ H2/2015.

 

[krumme]

---
I rather look out for 28XHP/20XHP and 28XLP/20XLP, than look at 14LPP/14LPe.

Where are you looking?

 

[krumme]

Galaxy S6 should be out by early April 2015, I think that's way too soon for 14nm FF. They only have two 20nm planar SoCs out there (Exynos 5430 and 5433) and those are available in specific models of the Galaxy Alpha and Galaxy Note 4 (most units are based on Snapdragon). I could be wrong but I think their first 14nm Soc will power either 2015 Galaxy Tab S line (they need a very powerful chip to compete with A8X/A9/A9X) or the Galaxy Note 5 @ H2/2015.

Thats my take too. If anything i think they will prioritize to get it into product where they can put pressure on qq prices. Besides getting the qq cheaper they can use qq to fill the gap so to speak. That is - if they have lower or higher yield they will just supplement with cheap qq stuff.
Ss will precisely know their own yields, qq - and tsmc- will not
...there is so much opportunity beeing the big guy its incredible.

 

[NostaSeronx]

Where are you looking?

Confidential documents from Samsung/GlobalFoundries. That some how get uploaded to scribd, and etc websites.

X = Fully Depleted/Extreme @ GlobalFoundries
28XHP/20XHP = 28nm/20nm GloFo/Samsung FDSOI
28XLP/20XLP = 28nm/20nm STM Licensed FDSOI
14XM/10XM = 14nm/10nm GloFo FinFETs

 

[Arachnotronic]

Thats my take too. If anything i think they will prioritize to get it into product where they can put pressure on qq prices. Besides getting the qq cheaper they can use qq to fill the gap so to speak. That is - if they have lower or higher yield they will just supplement with cheap qq stuff.
Ss will precisely know their own yields, qq - and tsmc- will not
...there is so much opportunity beeing the big guy its incredible.

I'm curious...why do you refer to Qualcomm as "qq"?

 

[geoxile]

The difference is GF cartoons was funny (at first). Samsung is kind of a dangerous brutal animal - with 1000 times more heft as backup.
The situation on the business side couldnt be more apart.

They're rich and powerful for sure, but you can't be that big and not suffer from the red tape of bureaucracy, especially in the context of Confucian hierarchies at the uppermost levels of management.

 

[MeldarthX]

They're rich and powerful for sure, but you can't be that big and not suffer from the red tape of bureaucracy, especially in the context of Confucian hierarchies at the uppermost levels of management.

Samsung doesn't really suffer from that; they are actually scary as they are one of the few mega corps that can and have turned their strat on a dime.....Company is ruled by an iron fist by the CEO.....people dragging their feet on something they get the boot; that includes the board....