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

[Techhog]

You are asking for something nobody can deliver because there isn't an official statement from Samsung or Apple. Just like you cant deliver the same with your 14LPE argument.

A9X is TSMC exclusive too. But its also a higher performance part.

"I'm correct because I know everything and I'm never wrong (except when I am)."

If you can't source something, don't state it as fact. That's something that you really struggle with (stating everything that you "know" as fact, even if you can't back it up).

 

[ShintaiDK]

"I'm correct because I know everything and I'm never wrong (except when I am)."

If you can't source something, don't state it as fact. That's something that you really struggle with (stating everything that you "know" as fact, even if you can't back it up).

You forgot to direct that at the LPE crowd too. Unless your intentions are one sided.

 

[ShintaiDK]

Apple's A9 is produced on 14LPE by both Samsung and Globalfoundries.

GloFo isn't producing anything for Apple.

 

[ShintaiDK]

Has anyone stopped to ponder,that BOTH Nvidia and AMD uarchs are very different and that different process nodes might actually work better for each of them. The AMD GCN cards run at a much lower sustained clockspeed than Maxwell and even to some degree Kepler did,so they will probably be going for lower clockspeeds and more shaders if the process is denser but is less efficient. Nvidia will probably go for higher sustained clockspeeds,with less shaders so a more efficient node will probably fit their way of doing things.

That could work out for them both as such. But we have to see. But its going to be pure speculation since it will most likely be Pascal vs some GCN 1.2/1.3. Will be interesting to see however, since the companies now go different ways in manufactoring.

 

[Fjodor2001]

"I'm correct because I know everything and I'm never wrong (except when I am)."

If you can't source something, don't state it as fact. That's something that you really struggle with (stating everything that you "know" as fact, even if you can't back it up).

+1. It's his usual MO. You can check my signature for a small selection of his previous "facts" that we just should accept as truth according to him.

 

[VR Enthusiast]

That could work out for them both as such. But we have to see. But its going to be pure speculation since it will most likely be Pascal vs some GCN 1.2/1.3. Will be interesting to see however, since the companies now go different ways in manufactoring.

There is no way to get to that conclusion alongside your previous conclusion that the A9 is LPP.

Assuming May 2015 for best-case production start on an LPP A9, that would mean 4 months from production start to iphones (September).

Arctic Islands is already ramping with M/P scheduled for February 2016. That puts Arctic Islands products at June 2016 latest, very likely to be lauched at Computex.

Pascal taped out in June 2015. There's no way it will take less than a year given what they are trying to do on a new process.

 

[Phynaz]

+1. It's his usual MO. You can check my signature for a small selection of his previous "facts" that we just should accept as truth according to him.

Five more days until my Intel quad core becomes slow and somehow costs me more ()🙂

 

[ShintaiDK]

There is no way to get to that conclusion alongside your previous conclusion that the A9 is LPP.

Assuming May 2015 for best-case production start on an LPP A9, that would mean 4 months from production start to iphones (September).

Arctic Islands is already ramping with M/P scheduled for February 2016. That puts Arctic Islands products at June 2016 latest, very likely to be lauched at Computex.

Pascal taped out in June 2015. There's no way it will take less than a year given what they are trying to do on a new process.

What uarch is Arctic Islands based on?

Also I doubt you see it anytime soon. Wccftech is a terrible "source"...

If it launched in June it would be GDDR5/GDDR5X or HBM1 based as flagship. But the Gemini X2 delay says it all. You know, Q2 product for a Fiji X2 card.

 

[ShintaiDK]

Five more days until my Intel quad core becomes slow and somehow costs me more ()🙂

Haha 😀

 

[VR Enthusiast]

What uarch is Arctic Islands based on?

Also I doubt you see it anytime soon. Wccftech is a terrible "source"...

If it launched in June it would be GDDR5/GDDR5X or HBM1 based as flagship. But the Gemini X2 delay says it all. You know, Q2 product for a Fiji X2 card.

I'm not using WCCFTech as a source for anything. It is clear that AMD successfully taped out multiple products on 14 LPP in early November and Global is on record stating that M/P starts early 2016.

There are other sources confirming this is true.

Fiji X2 is a niche product designed to market AMD's LiquidVR initiative. That is why it's "delayed". It will still be a very fast graphics card at a competitive price when compared to the new cards, and with lowered latency through Crossfire will be a good alternative option simply for VR vs a single Arctic Islands card. The "delay" also gives Crytek and Basemark more time to finish VRMark, which AMD wants to hit hard from the start. This one will raise eyebrows.

The HBM2 rumour sounds like you've been reading a bit too much of WCCFTech yourself.

 

[Abwx]

Where do you got this Abwx? Because it makes much sense on AMD dumping TSMC.

I already pointed this possibility but now i m 100% sure, because of what i quoted from the OP link :

http://english.etnews.com/20151222200002

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.

For a 3 years old process, if it s not sabotage i dont know what it is.

 

[Fjodor2001]

Five more days until my Intel quad core becomes slow and somehow costs me more ()🙂

No, ~1 year until 8 core Zen will obliterate your 4 core Intel CPU's in MT performance. And ∞ days until you will be able to comprehend content in forum threads.

 

[jpiniero]

Fiji X2 is a niche product designed to market AMD's LiquidVR initiative. That is why it's "delayed".

Considering how poorly Fiji has sold, I imagine they will probably just cancel the X2 at a later point. I don't think they would sell much if any.

 

[Maxima1]

But the Gemini X2 delay says it all. You know, Q2 product for a Fiji X2 card.

How do we know it's not just going to be canceled?

 

[JDG1980]

If the dual-Fiji card is released at all, I expect it will probably only be part of a promotional bundle with VR devices. That's the only application where it makes any sense; as a stand-alone card, who would want it?

I don't think we can assume anything about Arctic Islands release timing because of dual-Fiji's delay. AMD has said the delay was done to coincide with VR headset releases, and I see no reason to doubt that. Sure, they *could* be lying, but as noted, VR is about the only application where this card would be even remotely competitive, and we know AMD has had dual-Fiji prototypes kicking around for months; why would they be *unable* to release it now if they wanted?

 

[svenge]

We know AMD has had dual-Fiji prototypes kicking around for months; why would they be *unable* to release it now if they wanted?

Perhaps because their finances are so bad that their suppliers are requiring Net-0 credit terms?

 

[waltchan]

I'm confused with this article. Does this mean AMD have outsourced production to Samsung? Currently, who makes AMD AM3+ and FM2/FM2+ chips?

 

[ShintaiDK]

I'm confused with this article. Does this mean AMD have outsourced production to Samsung? Currently, who makes AMD AM3+ and FM2/FM2+ chips?

GlobalFoundries.

 

[Fjodor2001]

I'm confused with this article. Does this mean AMD have outsourced production to Samsung? Currently, who makes AMD AM3+ and FM2/FM2+ chips?

TSMC and Global Foundries produce chips for AMD currently.

According to the article, AMD's Next Gen Greenland GPU and Zen CPU will be produced by Samsung and Global Foundries on 14 nm process technology.

 

[maddie]

It seems that many are expecting a maximum die GPU as an early release.

Any other person finding it unrealistic to expect a Fiji sized GPU among the first 14nm GPUs released will realize that FuryX2 will be faster than any single Gpu AI card for the entire 2016. Only a full sized GPU will better FuryX2.

For VR no single 14nm GPU will outperform this card in 2016.

 

[ShintaiDK]

It seems that many are expecting a maximum die GPU as an early release.

Any other person finding it unrealistic to expect a Fiji sized GPU among the first 14nm GPUs released will realize that FuryX2 will be faster than any single Gpu AI card for the entire 2016. Only a full sized GPU will better FuryX2.

For VR no single 14nm GPU will outperform this card in 2016.

I dont expect big dies at all. GP100 may be the exception due to HPC sales. But gaming wise I think we gonna be lucky to even see 350-400mm2 as flagship. However nothing prevents a dual GPU 14/16nm card.

 

[DrMrLordX]

Considering how poorly Fiji has sold,

Have the Fiji sales been poor? They couldn't keep them on store shelves after launch.

 

[VR Enthusiast]

That's why the difference in die size matters here.

Nvidia's 500mm2 chip is AMD's 450mm2 chip on process difference alone.

AMD also needs to make use of the massive bandwidth they have left over with HBM.

I expect to see a 450mm2+ chip from them very early.

 

[jpiniero]

Have the Fiji sales been poor? They couldn't keep them on store shelves after launch.

Look at the Steam survey. The 300 series is the same way; the only AMD GPU that's sold decently in recent times is the 290/X when the price was severely cut.

 

[Fjodor2001]

I expect to see a 450mm2+ chip from them very early.

Are 14 nm yields good enough to allow for such huge dies? We've seen how poor the Intel 14 nm yields have been, despite that their process has been live for a long time now. And Intel's chips are much smaller than that. So is TSMC/GF/Samsung 14/16 nm yields that much better, so they can produce 450+ mm2 chips in 2016 at reasonable cost?