According to the insider source cited by the Korea electronic Times, AMD has decided to use Samsung for a portion of future GPU production. It hasn’t disowned GlobalFoundries, and will be splitting production of the new hardware between the two, as they share the same IP for that process. They famously began working together to create the new FinFet LPP process in 2014. That leaves TSMC out of the picture.
Production is expected to begin in Q2 2016, with Greenland GPUs coming first, followed by Zen CPUs shortly after.
This announcement of production has landed sooner than previously expected, which may be the reason that it is changing manufacturers. If TSMC has struggled with supply and yield issues, and AMD is targeting a specific date for launch, it can’t afford for any delays to happen because of fabrication issues.
It may also be that Samsung is simply ahead of the game in manufacturing these micro-part processes at this point. Greenland and Zen are both 14nm parts and we can expect that to be replaced shortly after by 10nm parts, possibly as soon as 2017 if some sources are to be believed. Samsung is said to already be developing its own fabrication process for that size of chip, so perhaps AMD wants to cosy up to it now, with a view to the future.
Of course, TSMC may not be overly worried. The fab has landed a portion of a lucrative contract with Apple, building chips for the iPhone 6s and iPad Pro.
Editors' Recommendations
- I’ve used Intel CPUs for years. Here’s why I’m finally switching to AMD
- AMD Zen 5: Everything we know about AMD’s next-gen CPUs
- Intel may fire the first shots in the next-gen GPU war
- AMD’s next version of FSR promises better visuals and support for Xbox
- AMD needs to fix this one problem with its next-gen GPUs
