AMD Likely Skipping 10nm For 7nm, Restructures GlobalFoundries

top kek always behind the curve as usual

File: 1472056418712.jpg (113KB, 960x960px) Image search: [Google]

113KB, 960x960px

AMD is pretty much over.

>>56380248
so we are not going to see a new ayyymd processor never again.

sad news indeed.

>Paying GloFo $100M penalty for shitty process
>Paying GloFo for using other fabs

TOP KEK

>>56380291

MOAR!

File: tidus.jpg (30KB, 475x343px) Image search: [Google]

30KB, 475x343px

>>56380248
>3dfx roaldmap beyond voodoo 6k

>>56380291
AMD has GPU loli

>>56380248
When Global Foundries complete the deal with IBM to absorb their foundry business GloFo got IBM's wealth of process IP, including two completely developed nodes, and the right to license those out for their customers.
IBM has a high performance 14nm SOI based FinFET node used for their own POWER9 chips now being fabbed at GloFo.
IBM also developed a 7nm node which will be the bleeding edge process brought online end of next year.
GloFo will not offer a 10nm node like competing foundries, they will instead skip to 7nm.

AMD amended their wafer supply agreement with GloFo in order to have access to each of these.
While it would be technically possible for Zen+ based chips to launch on this 7nm process, AMD will likely leverage 14nm SOI as their next node for decreasing power and gaining frequency.

>>56380315
No one should ever make any mistake, GloFo is basically run like a Mafia. AMD is their only consistent big customer and they're still run through the ringer.

GO AMD GO

>>56380958
down the loo

>>56380939
The new agreement also gave them less penalties to use other competitor's fabs, which means samgsung's and tsmc's 10nm probably. Although does GF and sammy's tech share not extend to 10nm?

On another note, is Samsung's 10nm just going to be a 14nm physical node with finfet? That would put IBM's (GF) 14nm soiFF on the same feature size at least.

>>56381010
Samsung and GloFo didn't collaborate on a 10nm process, their agreement only covers the bulk 14nm LPP process.
No cohesive specifics on IBM's 14nm SOI process are available publicly that I can find, so I couldn't make a comparison on area scaling. Maybe they'll talk about it at IEDM. They have detailed a lot of technical stuff on the research that went into building SOI FinFETs though I don't know if any of it is relevant to the exact metrics of their commercial ready node.

>>56380248
kek
Can't even into 10nm
And they think they can jump down to 7nm
AMD E A D

>>56381274
20nm was such a disaster pretty much everyone skipped it. 10 could have similar issues

>>56382734
20nm was a planar shrink that had horrible leakage, and 14/16nm is basically 20nm plus FinFET to address that.

10nm is supposed to be an actual shrink again and it wouldn't surprise me if 7nm is just that same geometry again plus whatever tweaks are needed to address the resulting problems.

> fuck these hebrew-tier naming conventions in the semiconductor industry. just tell me the geometric mean of the gate and metal pitches so I at least have a reasonable density comparison

>>56383025
Why are you conflating IBM's process technology with conventional bulk solutions from Samsung and TSMC?