Global Foundries has announced that they are finally making chips in their new Fab 8 in New York, but there is a surprise twist. That twist is IBM is now jointly making chips at their New York fabs too.
The current production parts are 32nm HKMG SOI on 300mm wafers, so that leaves two possible candidates for the customer. Since this is a release with IBM Microelectronics, that customer isn’t likely AMD. In any case, the production is currently small, with volume ramping up to full capacity by the second half of the year.
There are a few things to note here, timing being the key one. The production start is a bit late, likely less than a quarter, but given the cost of a fab, there are a lot of zeroes in that 3 months. This could have a lot to do with the recent departures of several Fab 8 executives, but this hasn’t been stated directly.
Next up is the ramp. What you will see between “first full-flow customer silicon” and “full production” is basically taking delivery of tools, putting them in place, and calibrating them. This is a painstaking and detail oriented process, and one that can’t really be rushed. If they can produce finished silicon, that means the bugs are going to be minor and likely related to a specific machine from here on out.
Last up is the process itself, and that is a bit of a surprise. 32nm SOI is not really a high demand process, but it is one that AMD in particular needs more capacity on. Given that 28nm is way behind schedule, it is somewhat surprising that GloFo didn’t start there. This could be because of demand, or more likely to start with a known good process rather than one still in the debugging phase.
In any case, both GloFo and IBM are now making 32nm HKMG silicon in New York. The Fab 8 is running, and the big hurdles appear to be a thing of the past. Any bets on when 28nm customer silicon will make it out the door in upstate New York?S|A
Latest posts by Charlie Demerjian (see all)
- AMD puts massive SSDs on GPUs and calls it SSG - Jul 25, 2016
- AMD revamps workstation lines and adds three cards - Jul 25, 2016
- AMD unveils Loom and ProRender software - Jul 25, 2016
- Emulex puts NVMe over FC with NVMf - Jul 20, 2016
- ARM needed Softbank, Softbank wanted ARM - Jul 19, 2016