Major semiconductor manufacturer TSMC has revealed its plans for a 12nm process node. TSMC hopes to roll out the 12nm fabrication process as an advanced version of its current 16nm technology found in many of Nvidia’s Pascal GPUs.
TSMC has been a superpower in the semiconductor industry for a number of years now, and dominant when it comes to the 28nm process segment, but has found itself left behind with the competing technologies coming out of GlobalFoundries and Samsung. The 12nm process node is intended to help TSMC compete with other lower process nodes.
The revised process node will essentially be a smaller version of its current 16nm FinFET process coupled with a lower cost and improved leakage characteristics (less performance wasted).
While the different between 16nm and 12nm at first seems small, it's estimated the 12nm process node will offer around 15% faster performance despite 50% lower power consumption, all at a cheaper cost than current 16nm chips. It sounds like a win/win and it’s only a matter of time until Nvidia and AMD look to take advantage of 12nm process nodes. Currently AMD’s RX 400 Polaris series is manufactured using 14nm GlobalFoundries while Nvidia Pascal is 16nm TSMC.
Could we perhaps see revised Polaris and Pascal chips using TSMC’s new process in 2017? If we do then the leap in performance is going to be evolutionary rather than revolutionary. It’s not every year we can get a Pascal-esque jump in performance.
TSMC is hoping it can begin to generate revenues from the 12nm process node from early 2017 so it appears as if manufacturing contracts are already in the pipeline. The first hardware to roll out should be with us later in 2017.