[AMD] K12 will be on 28nm

[mrmt]

http://seekingalpha.com/article/273...nductors-software-and-supply-chain-conference

(...)

Hans Mosesmann - Raymond James
Okay. And so the K12, the custom version is 2016…

Devinder Kumar - SVP and CFO
2016, that’s right.

Hans Mosesmann - Raymond James
28 nanometer for…

Devinder Kumar - SVP and CFO
28, not yeah, I think it will be 28 nanometer, but you know it’s more from my standpoint the products that I think you want to focus on in terms of what the products can do, you know we be a process technology as necessary but you don’t have a leading edge to these days to go ahead and give the kind of performance or the path that we introduced.

So K12 will be on 28nm and there isn't a single mention to Zen in the last two conferences. It seems that K12 and Zen are indeed low cost, scope constrained projects that won't have any relevance at all on the consumer and server market.

ed: link included

 

[witeken]

So 28nm will go into its 5th year as mainstream foundry node?

 

[Sweepr]

What a joke. Samsung already has two 20nm HKMG SoCs out and the first 14nm FinFet SoC could be introduced as soon as Q2 next year (Galaxy S6 launch). How does AMD plans to keep their products relevant by launching a brand new architecture in an outdated process?

Also:

AMD’s Amur SoC will be targeted dead center at the android ecosystem, specifically the tablet market and will probably feature 4 cores. Its graphics part will be powered by Graphic Core Next (probably 2.0). Nolan on the other hand is based on the Puma+ architecture and will be virtually identical to its ARM counterpart apart from the fact that it is x86 and not ARM. Both Nolan and Amur are slated to be produced on the 20nm node and have a release time stamp of Q3 2015.

This rumour came out in October. Two weeks ago we got an official response from AMD (Mobility Roadmap):

No 20nm x86 Nolan SoC in 2015 (Carrizo-L is still 28nm), no word on 20nm ARM Amur SoC either.

 

[witeken]

How can you release a high volume product with an SoC that's built on a node that has yet to go into HVM? (http://forums.anandtech.com/showpost.php?p=36944103&postcount=68)

 

[Arachnotronic]

So K12 will be on 28nm and there isn't a single mention to Zen in the last two conferences. It seems that K12 and Zen are indeed low cost, scope constrained projects that won't have any relevance at all on the consumer and server market.

Oh dear...surely this is a case of the CFO simply misspeaking?

28nm in 2016+ is just not going to be competitive.

 

[liahos1]

28nm in 2016 wow. somebody needs to update wiki on it.

"some 20nm" product in 2015. sort of strategically ambiguous on when they introduce a product with finfet.

i like how he plays down the importance of process node.

 

[Sweepr]

How can you release a high volume product with an SoC that's built on a node that has yet to go into HVM? (http://forums.anandtech.com/showpost.php?p=36944103&postcount=68)

Don't ask me. That info comes from Andrei Frumusanu, he seems to have some insider knowledge about Samsung products. I personally don't believe we will see a 14nm SoC from Samsung till the Galaxy Note 5 (late Q3/2015). Also:

We have started mass production wafers in 14-nanometer. So, in 14-nanometer, our progress is well on track, including process and the yields. Here we're talking when I mention yields being on track, this is for a very advanced product. It'll be our most advanced applications processor. Yielding at full specifications targets with all of the characteristics for speed and power as well.

 

[geoxile]

That's rather strange. Perhaps Zen and K12 will be significantly bigger than initially thought and built on advanced FD-SOI?

 

[waffleironhead]

When I read that, I get him saying he doesn't really know.

"28, not yeah, I think it will be..."

 

[rtsurfer]

Fail AMD.

Common guys, 28nm in 2016.

Good luck getting anywhere near Intel in Power Consumption.

Performance is another concern.

 

[NTMBK]

Got to be a mistake. If that is true, it's utterly DOA.

EDIT: A link to the source would be appreciated, thanks!

 

[DeathReborn]

If true it puts that bleak future article further into prominence. If they can't even get the ARM version onto 20nm/16nm/14nm in 2016 the writing would be on the wall for their phone/tablet hopes.

 

[geoxile]

If true it puts that bleak future article further into prominence. If they can't even get the ARM version onto 20nm/16nm/14nm in 2016 the writing would be on the wall for their phone/tablet hopes.

They said K12 is for embedded, dense server, and thin client. While those are low power that can be higher power than ultra mobile like phones and tablets.

In light of this news I think it's more likely that K12 and Zen will be large cores.

 

[Burpo]

http://www.investorvillage.com/smbd.asp?mb=476&mn=293826&pt=msg&mid=14446382

"Devinder Kumar - SVP and CFO
Today the bulk of our product is 28 nanometers at both foundries. We will have certain products move to 20 nanometers in 2015. And then, going forward we will be in FinFET timeframe to be determine based on product intersection with technology needs on a go forward standpoint. But we will get FinFET. Now the thing that’s interesting that you might have observed is GlobalFoundries signed an alliance agreement on future technology nodes including FinFET with Samsung and very recently they’ve gone ahead and had the arrangement with IBM, whereby they’re going to take the Microelectronics division. So that’s already helpful for us because it gives them the scale, gives them additional customers and from all those standpoint a healthy and GlobalFoundries as doing well is obviously very good for AMD."
Original Source..
http://seekingalpha.com/article/273...tware-and-supply-chain-conference?part=single

 

[Abwx]

It seems that K12 and Zen are indeed low cost, scope constrained projects that won't have any relevance at all on the consumer and server market.

Where is the source, any link.??.

As for the rest of your post that s obviously BS if we are to read what AMD s enginering think about their K12.

Papermaster declined to go into detail about K12, saying simply that it is a “state of the art of CPU design for those tasks such that you need extensive power management wherever you go — even in a server, not just on the laptop,

I ended the discussion with a question about K12. While Moshkelani also was not about to disclose intimate details, he smiled when I mentioned other custom CPU cores, such as Qualcomm‘s “Krait,”

Moshkelani said there will be a significant difference in power efficiency and performance to the advantage of K12.

http://blogs.barrons.com/techtraderdaily/tag/server/page/4/

 

[Exophase]

In light of this news I think it's more likely that K12 and Zen will be large cores.

On 28nm they'll be large all right...

 

[Abwx]

I m not fluent enough in english, can someone give explanations about the meaning of this sentence :

28, not yeah, I think it will be 28 nanometer, but you know it’s more from my standpoint the products that I think you want to focus on in terms of what the products can do

So is it no or is it yes..??.

 

[ShintaiDK]

So 28nm will go into its 5th year as mainstream foundry node?

That was expected due to cost.

 

[ShintaiDK]

So is it no or is it yes..??.

He basicly defends the usage of 28nm and says whats more important is how the product functions and not what node its made on.

 

[Lepton87]

I m not fluent enough in english, can someone give explanations about the meaning of this sentence :



So is it no or is it yes..??.

It means he's not sure but he's of the opinion that it's going to be manufactured in 28nm process and then he proceeds to downplay the process node as an important product characteristics. But it was very clumsily worded, someone higher than a cleaning lady in an IT company shouldn't use such language.

 

[Abwx]

Anyway if false we ll get a statement by AMD aknowledging the mispeaking, and if true then they ll say nothing since it would already have been told...

 

[mrmt]

Got to be a mistake. If that is true, it's utterly DOA.

I don't think Kumar would be *that* amateurish.

http://wsw.com/webcast/rj94/amd/

check around 13:53

 

[Arachnotronic]

I don't think Kumar would be *that* amateurish.

http://wsw.com/webcast/rj94/amd/

check around 13:53

Yup, it's 28nm. He wouldn't have downplayed process node if they had a 14nm part waiting in the wings for 2016

 

[WittyRemark]

I'm not so sure how I feel about this.

 

[meloz]

The reason they might be 'stuck' at 28nm is economics: AMD likely cannot afford to bid for expensive wafers on tighter nodes at TSMC (or Samsung, if we are daydreaming), and must wait until it becomes cheaper and more accessible to them via Global Flounderies. They cannot afford those expensive wafers because they know they cannot sell the end products (CPUs/APUs/SoCs) at a price to justify the expense of 20nm and narrower manufacturing.

Hopefully this news comes as a much needed wake-up call for those deluded souls who believe that merely by building a CPU/APU around some ARM architecture AMD will be back in the game and slay Intel.