AMD Announces Multi-Year Amendment to the Wafer Supply Agreement With GLOBALFOUNDRIES

[superstition]

Maybe not "difficult" (hint - all this stuff is difficult), but very expensive and time consuming. What good would a 16nm Zen be two years from now?

Considering that Apple already did it the infrastructure/procedural knowledge should be in place, making it easier for AMD to accomplish the same or similar.

 

[Elixer]

The only negative here is the dilution of stock for the stockholders, which is why the stock took a slight dip on this news, but it's recovered since. Other than that this looks positive. In fact it looks to me like AMD are convinced they might need more capacity than GloFo can provide on 14nm and they want to be able to dual source, when was the last time AMD were worried that they couldn't make enough CPUs to satisfy demand?

Any talk regarding the figures involved is conjecture. Since we have no clue how many wafers AMD is getting for that money or other details of the agreement.

If I were Intel I'd be worried.

What exactly would Intel need to be worried about?
Intel has been in a holding pattern for a very long time, they are sitting on boat loads of cash, and, they could undercut AMD in a heartbeat if they wanted.

This all seems to hedge on, how many orders they got, and for what.
We know Sony & MS chips are built at TSMC, Polaris & APUs and all other chips that we know of at GloFlo.
We also know, under the old contract, they could use other fabs. Now, they seem to be paying GloFlo for using other fabs, so, it don't seem like that good of a deal at all.

 

[VirtualLarry]

We also know, under the old contract, they could use other fabs. Now, they seem to be paying GloFlo for using other fabs, so, it don't seem like that good of a deal at all.

I thought that in a recent past WSA modification agreement, that AMD had agreed to exclusivity with GF. (Probably on promises of their license 14nm process tech.)

Obviously, that didn't work out so well, so they wanted a way out of the exclusivity.

 

[sirmo]

What exactly would Intel need to be worried about?
Intel has been in a holding pattern for a very long time, they are sitting on boat loads of cash, and, they could undercut AMD in a heartbeat if they wanted.

Running all those fabs costs money. It's highly likely they will lose marketshare from AMD. Perhaps laying off 11% of its staff was in anticipation of AMD's comeback to the x86 market.

 

[Phynaz]

Considering that Apple already did it the infrastructure/procedural knowledge should be in place, making it easier for AMD to accomplish the same or similar.

Apple did NOT do it. What Apple did was design two CPU's, one that is fabbed at TSMC and one that is fabbed at Samsung.

You can't just take a chip design and send it over to an unrelated foundry, the design rules are different.

 

[superstition]

Apple did NOT do it. What Apple did was design two CPU's, one that is fabbed at TSMC and one that is fabbed at Samsung.

You can't just take a chip design and send it over to an unrelated foundry, the design rules are different.

The design rule overlap should be sufficient to make it practical. It was for Apple. That was the point I apparently didn't make clearly. I am familiar with the design rule issue. TSMC's guy said the number of rules proliferates as nodes shrink, for instance.

 

[Phynaz]

Design rule overlap? Now you're just making things up.

If it's so easy, why is GF having such problems at 14nm when it's Samsung's manufacturing process. Even when the "rules" are the same it's difficult to implement them.

But you didn't answer the question, what good would it do to have new Zen CPU out two years from now? AMD has enough problems with late launches without making it intentional.