TSMC’s 10nm FinFET Process Is On-Track – Ready To Begin Ramp to Trial Production at Fab 15 in 2H 2016
Taiwan Semiconductor Manufacturing Company (TSMC) is one of the leading pure play foundries of the world and its operations are of serious interest to PC hardware enthusiasts. The company is currently working on the 16nm FinFET process but has stated in a press release that its 10nm dabblings remain on track and on schedule. The company has broken ground for a new 10nm Fab in June and plans to ramp upto trial production in 2H 2016.
TSMC reiterates that its on time for 10nm, trial production to begin at Fab 15 in 2016
A few months ago, the company stated that it will be breaking ground for a new fabrication plant dedicated to the 10nm process in June. The facility will be named Fab 15 and will be based on 300mm wafers. The company has been loosing orders to Samsung (which is brute forcing yield problems by having deep pockets) and this is one attempt of many to try and get their strong position with Apple and Qualcomm back. Graphic cards on the other hand, currently remain a TSMC business. The company has previously urged customers to free up its 200mm capacity as well - the reason for which currently remains unknown.
If production begins in 2H 2016 as planned then the first products should hit the shelves by 2017. TSMC’s 10nm FinFET (CLN10FF) will have 110 – 120 per cent higher transistor density compared to its 16nm FinFET+ (CLN16FF+) process tech, 15 per cent higher frequency potential at the same power and 35 per cent lower power consumption at the same frequency and complexity. Something worth noting is the fact that the characteristics of the node have not been finalized although one can make a pretty good guess as to its physical traits.
TSMC’s 16nmFF is actually just 20nm with FinFETs so 10nmFF should be on par with Intel’s 14nm node. Ofcourse as I have stated many times before the market won’t care about the quality of the chips. They won’t really look at the Gate pitch & Fin Pitch, and this could adversely affect Intel Corporation which usually churns out completely clean nodes.
TSMC has also released plans to ramp 7nm somewhere in 2018 with “10nm elements”, so we already know that TSMC’s 7nm should be based on the 10nm node in general. Intel, the world’s only company that actually churns out true nodes in every aspect with more or less perfect scaling, has not released concrete plans for 7nm. Infact it is even contemplating material changes, although EUV will be a must since it can easily fabricate chips around the 7nm mark without use of double patterning.
They are employing a task force consisting of 300-400 people for R&D of the 10nm Node. Previously, the majority (around 85%) of the production capability was employed in 28nm Node, specifically the HKMG (High-k Metal Gate), which is now transitioning gradually to the 16nm FinFET process. If the current situation is any indicator then the current 10nm plans will only be for non-GPU chips until the process matures to the point where it can support high performance ASICs, most probably by 2017.