Samsung and Global Foundries will produce AMD's Next Gen Greenland GPU and Zen CPU

[Fjodor2001]

http://english.etnews.com/20151222200002

Samsung Electronics Does Toll Manufacturing for AMD’s Next Chip

Samsung Electronics secured AMD as its toll manufacturing (foundry) customer for semiconductor chip and will produce AMD’s next graphic processing unit (GPU) chip starting from New Year. After securing AMD as its customer along with Apple and Qualcomm, Samsung Electronics’ project on making its foundry world-class has become faster. AMD is a business that had controlled over PC CPU markets along with Intel and it is also the number 2 business in GPU market after NVIDIA.

According to industries on the 21st, Samsung Electronics’ System LSI Foundry Business Department will start mass-producing AMD’s new GPU ‘Greenland (development code name)’ along with Global Foundry (GF) starting from 2nd quarter of 2016.
[...]

Proportion of production supplies is very fluid as AMD will regulate proportion between 2 businesses according to many conditions such as yield and others.

AMD had been asking TSMC from Taiwan to be in charge of GPU production but it decided to cut contract with TSMC with 28-nano being the last production after issues with yield and instable supplies had continued.
[...]

“Because Samsung Electronics and GF have same IP for 14-nano processing, chips that are designed by AMD will all be produced at both factories.” said a person who is familiar with this industry. “If products are produced from both factories, AMD won’t have to worry about a problem regarding lack of supplies.”
After starting production of Greenland, AMD will put out its new CPU ‘Zen’ right after. This product is also produced from 14-nano LPP processing and will be produced by Samsung Electronics and GF just like GPU.
[...]

Samsung Electronics will speed up process of development of 10-nano processing that will be regularized in 2017. It will make an announcement of its thesis based on successful development of S-RAM that will be loaded into 10-Nano Logic at ISSCC Society that will be held in January of 2016 and do early promotion about it.

So it will be Samsung and Global Foundries producing AMD's next generation Greenland GPU and Zen CPU after all. TSMC is out of the game.

Samsung also seems very aggressive on moving ahead to 10 nm.

 

[ShintaiDK]

Apple on the other hand dumped Samsung and GloFo in favour of 16FF+ with the A10.

 

[jpiniero]

Apple on the other hand dumped Samsung and GloFo in favour of 16FF+ with the A10.

On the A9, Samsung's chip is 8% smaller but TSMC's draws less and I think is even a bit cheaper. Not sure how that would play out in a very large GPU though.

 

[SlowSpyder]

AMD can be competitive already in some workloads with an old design and ancient manufacturing process compared to Intel. Give them a state of the art manufacturing process and they may be able to put out some pretty good silicon, looking forward to see what products they have for sale in the not too distant future.

 

[Tuna-Fish]

My google-fu is failing me, is the Samsung Apple A9 LPE or LPP?

AMD on Samsung/GF and nVidia on TSMC is interesting if nothing else. I wonder how much difference there is between the nodes, really?

 

[ShintaiDK]

My google-fu is failing me, is the Samsung Apple A9 LPE or LPP?

LPP.

 

[fourdegrees11]

AMD definitely needed to move with a fab that can actually deliver a product on a new node, on time!

 

[beginner99]

I guess they went with Samsung so that as soon as GF finally gets their shit together they can move there due to WSA.

 

[mrmt]

If this is true I wonder if there is some kind of agreement between AMD and Globalfoundries so every order placed to Samsung counts towards fulfilling the WSA quota. With AMD's ever-shrinking pool of orders this is a very important issue.

 

[William Gaatjes]

Very good. I am pleased to read that AMD has secured production.
Now all we have to do is wait.
And hopefully AMD will rise again. The other alternative will be that AMD will be split up again and again until nothing is left.

I do think that this partnership with samsung is good for technologies. HSA being one of them. If all goes well, partnership between Samsung and AMD for future versions of HSA might happen and be a good thing for the x86-64 and ARM platforms. I would not be surprised if it is AMD who announces another new feature set for the x86 cpu. One that tightly integrates HSA. Just as they did with the x86 64 bit platform. With a 64bit address space, cpu and gpu and any other dedicated computing unit can have all the memory they want in a single 64 bit linear addressing space. the MMU will take care of data protection.

And that is just AMD and Samsung. There are a lot of huge names in the HSA foundation.

http://www.hsafoundation.com/

 

[Borealis7]

Sh*t just got REAL!

I hope we would be able to see the results in 2016 in the GPU scene (i doubt Zen will be a 2016 product).
but something i don't understand...Samsung will make chips on the LPP node, but aren't GPUs the exact opposite of a "low power" product?

EDIT: okay, Russian explained it on the VC&G thread (http://forums.anandtech.com/showpost.php?p=37916859&postcount=15)

 

[dark zero]

Samsung doing AMD chips? HALF LIFE 3 CONFIRMED...

ok, no, but being serious Samsung has better reputation than Glofo. Definately they will do the top tier chips, while GloFo will do the low and mid tier chips.

And we see where HSA is heading.. to Android platform

 

[moonbogg]

In another thread people talked about AMD going under and who would make a good replacement. I said Samsung could make their chips and would be a good replacement (red tape be damned). People had technical reasons for saying I was crazy, but they are the ones who now have egg running down their face, tiny bits of shell stuck in their eye brows, and yoke just went up their nose somehow.
Also, AMD GPUs will be on a smaller process than Nvidia's. I expect AMD to have an efficiency advantage and it will shake things up like the craziness.

 

[mrmt]

Also, AMD GPUs will be on a smaller process than Nvidia's. I expect AMD to have an efficiency advantage and it will shake things up like the craziness.

If anything Samsung 14nm is less efficient than TSMC 16nm, it will be quite a surprise if AMD suddenly becomes more efficient to the point of overcoming Nvidia's architectural advantages.

This, together with the predicted lackluster performance of Zen on HPC and the fact that they are still playing catching up on the software ecosystem should put AMD HPC ambitions to rest.

 

[Arachnotronic]

If anything Samsung 14nm is less efficient than TSMC 16nm

Correct. There is a reason why Apple is switching away from Samsung 14nm completely in the coming iPhone cycle.

 

[maddie]

In another thread people talked about AMD going under and who would make a good replacement. I said Samsung could make their chips and would be a good replacement (red tape be damned). People had technical reasons for saying I was crazy, but they are the ones who now have egg running down their face, tiny bits of shell stuck in their eye brows, and yoke just went up their nose somehow.
Also, AMD GPUs will be on a smaller process than Nvidia's. I expect AMD to have an efficiency advantage and it will shake things up like the craziness.

You are very optimistic. The ones who criticized you probably loves egg and shell bits. Nothing changes. Rabid posts are the norm.

 

[ShintaiDK]

On the A9, Samsung's chip is 8% smaller but TSMC's draws less and I think is even a bit cheaper. Not sure how that would play out in a very large GPU though.

If it is like previous. The bigger but more effective wins

 

[moonbogg]

If anything Samsung 14nm is less efficient than TSMC 16nm, it will be quite a surprise if AMD suddenly becomes more efficient to the point of overcoming Nvidia's architectural advantages.

This, together with the predicted lackluster performance of Zen on HPC and the fact that they are still playing catching up on the software ecosystem should put AMD HPC ambitions to rest.

I read that, but it was after my post about eggs and stuff. Well, looks like I really blew it then. I thought I could outsmart you tech nerds by pointing out the obvious advantages of a smaller MFG process. But of course, there was something unusual and different about this particular smaller process that somehow made it less efficient than a larger process, and of course you guys knew about that somehow. I feel like I'm a villain in Scooby Doo right now. How the hell you guys do that?

 

[USER8000]

predicted lackluster performance of Zen on HPC

Where was this prediction - you got the documents to share here,so we can all see them??

I read that, but it was after my post about eggs and stuff. Well, looks like I really blew it then. I thought I could outsmart you tech nerds by pointing out the obvious advantages of a smaller MFG process. But of course, there was something unusual and different about this particular smaller process that somehow made it less efficient than a larger process, and of course you guys knew about that somehow. I feel like I'm a villain in Scooby Doo right now. How the hell you guys do that?

If its supposedly more denser and consumes more power,they can pack more logic into the chip and run the core at a lower clockspeed to compensate.

The Apple comparison is only valid since its exactly the same chip running at the same clockspeeds - its not the same with AMD and Nvidia,with the former running their cores at significantly lower clockspeed than the latter.

But then,if Apple is using a lot of the volume from TSMC currently,it means Samsung has a whole lot of volume free,meaning AMD can release chips earlier than Nvidia too,who will be fighting for orders from Apple.

This might be not such an issue for the lower volume,highest end chips,but it makes me wonder about the much larger volume midrange and lower highend chips.

 

[prtskg]

LPP.

http://www.tomshardware.com/news/iphone-6s-a9-samsung-vs-tsmc,30306.html
It was 14nm lpe. Neither lpp nor finfet+ reached volume production when A9 came. We'll see lpp products in 2016.

 

[Arachnotronic]

http://www.tomshardware.com/news/iphone-6s-a9-samsung-vs-tsmc,30306.html
It was 14nm lpe. Neither lpp nor finfet+ reached volume production when A9 came. We'll see lpp products in 2016.

Tom's Hardware is incorrect. A9 is LPP.

 

[VR Enthusiast]

Tom's Hardware is incorrect. A9 is LPP.

Do you have a link to proof of this? I'm not even sure Samsung has LPP at all otherwise AMD would be using them earlier than GlobalFoundries instead of later, wouldn't they?

 

[prtskg]

Tom's Hardware is incorrect. A9 is LPP.

Toms is right as neither lpp nor finfet+ reached volume production then.
Samsung itself thinks lpp is important for 2016 and after (not 2015 and after)-
http://www.samsung.com/semiconductor/foundry/process-technology/14nm/
TSMC 16nm finfet (not finfet+) reached volume production in Q2 2015
http://techreport.com/news/28662/tsmc-begins-volume-production-of-16-nm-finfet-chips

 

[Andrei.]

Tom's Hardware is incorrect. A9 is LPP.

It's LPE.

 

[Arachnotronic]

It's LPE.

That's not what several independent industry sources have told me.

As the story goes, Samsung basically used A9 as its yield learning vehicle for 14LPP (Exynos 7420 was used for LPE).