AMD and Globalfoundries are announcing the birth of their new 14LPP process, cigars from SemiAccurate all around. Actually the big news really is that the 14nm process has gotten fully qualified with a big customer, AMD, and working silicon has been delivered.
It was pretty clear a long time ago that AMD was skipping the 20nm process, that is why the Zen SoCs/CPUs are delayed. Last spring the plan was changed to scrap the woeful 20nm process and refocus on 14nm. With the self-inflicted wound of bad messaging, few outlets actually understood this process enough to not cry that the sky falling.
So now with 20nm rightly shelved and 14nm FinFET based parts imminent, AMD and Globalfoundries are announcing that the process is all ready to go. Normally we would be happy to hear this but you might recall that the GF 14nm process is a transplanted Samsung process. Samsung was producing volume products, Apple A9 volumes as SemiAccurate exclusively reported last December. Globalfoundries isn’t going to have their ramp at AMD volumes running at full steam until some time in 2016.
Why it took over a year for Globalfoundries to get up to speed on a finished 14nm process is beyond us but they are there. With any luck the ramp from here on out will be smooth and trouble-free, both AMD and GF need a break. I guess we will know when the release product for the current working silicon becomes official. That said AMD did not say that the working silicon delivered was a product, just that it was silicon, made on the 14LPP process, and was delivered presumably to them.
It all seems rather vague and non-informational but that is by intent. Luckily we can read some things into the information and narrow down the times considerably. Lets start with an assumption that the silicon delivered was delivered a few weeks ago, enough time for the jolly elves of Sunnyvale to run the appropriate tests and give the silicon a thumbs up. Getting chips back and going public before you know they or the process they are built on works and works at volume is a bad idea. Some go as far as to wave their failure in front of hundreds of press, analysts, and a few unwashed too. We strongly recommend the conservative route though.
Note: The following is for professional and student level subscribers.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Latest posts by Charlie Demerjian (see all)
- HyperX ships it’s 60 millionth enthusiast memory module - Oct 15, 2018
- Bittware/Nallatech water cools 300W of Xilinx FPGA - Oct 12, 2018
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018