Samsung 10nm FinFET Process Scheduled to Start Trial Production in late 2015 – Volume Ramp in 2016
Samsung Foundry has revealed (via KitGuru) that it will begin volume production of its 10nm FinFET node in late 2016 (with the chips hitting the shelves sometime in 2017). We have also previously heard from TSMC about stepping up their production to 10nm in 2016 as well. This leaves only Intel, which is slated to start production somewhere late 2016 to early 2017 with products coming in 2H 2017.
Marketing material from Samsung. @Samsung Public Domain
Samsung adds 10 nanometer to the roadmap, ramps to volume in 2016
Samsung has stated that the process will have smaller transistor fin, transistor gate and interconnect pitches smaller than the existing 14nm node, but if historical data is any indication, Intel will be the only company to have a true 10nm node in this time frame. The shrinking of these critical factors is what constitutes a process upgrade since it allows for higher density and a more cost effective process. Samsung had the following to say on the occasion:
“We have a full PDK [process development kit] and silicon-verified IP available, and are, currently, running MPWs [multi-project wafer] now for our lead customers,” With mass production scheduled from late 2016 we are excited about this next phase of our foundry business. Early designs activities have already sparked some interesting discussions.” via KitGuru
Samsung has previously demonstrated a 300mm wafer that was fabricated on its 10nm technology. Trial production is scheduled to begin sometime in late 2015, while volume production for the commercial chips will begin in 2016. The more closer Samsung is to a true 10nm process, the more yield problems it will face if it ramps prematurely. However, it being Samsung, it has the option to simply brute force low yields by spending more capital - something it could choose to do to undermine Intel's lead.
TSMC’s 16nmFF is actually just 20nm with FinFETs so 10nmFF should be on par with Intel’s 14nm node. However, both Samsung and TSMC nodes have not completed the specification phase, so it is premature to say how much exactly. But as I have stated many times before the market won’t care about the quality of the chips. It is doubtful that they will look at the technological differences between TSMCs/Samsung's 10nm a true 10nm node - rather see Intel's lead diminish.