AMD has renegotiated its deal with GlobalFoundries. From now on the chip designer will only pay GlobalFoundries for 32nm dies that work instead of a fixed price per wafer. Full details in the PR below.... More...
i would think this will drive overall costs up as the failure rate expense will be passed on into the chips that pass.Guess we will see
If GlobalFoundries has other customers....the cost will more than likely be passed onto those customers.
That's not true at all, other designers like Broadcom, Qualcomm and STMicroelectronics uses GlobalFoundries' fabs. deltatux
We need AMD to be up and running or Intel will charge us more and more for their products. The prices of Intel hardware in my country are just hilarious.
don't think it happen . If Global foundries is anything like TSMC they have two sort of pricing Per chip or per wafer and it's up to you to choose which model to get. by the way for NVIDIA when they had less than 20% yield on fermi chips It was still more profitable to pay per wafer , so this arrangement raise the question if there is some serious problem with GLOFO 32nm Line. or perhaps problem lies with the chip that AMD asked them to build
I see nothing rude about telling someone you'll only pay for parts that actually work. It's like telling a farmer you'll only pay for produce that isn't rotten.
Considering that it's literally impossible not to have defective ones, in fact, it's basically required...to not pay them for that, I just see it as a dick move. It's not like farming where if you do it right, you get your whole crop...with this type of thing you're expected to have like a 20% defective rate...when they purchase them, they know this.
The problem is that they don't have the incentives to try improving the yield rate if they pay for the defectives as well so the income is guaranteed no matter how bad the yield.
Ye this should put pressure on GlobalFoundries to improve yields because GlobalFoundries has already delayed 32nm for a while. deltatux
Not something you can just improve. It's just how it works. You can only get so much out of something.
Its not something that can just improve but something that they can avoid to messed up and increase the average yield, it is also a incentive to revise the procedure they use to avoid human errors.
No, they take what is left-over from the pure ones, stick em together and hope they work...these are the ones that fail usually. Basically, this will just make the company stop producing as much as they do...so there will be very 32nm chips on the market in the long run.