Hey guys, I saw a couple people talking about 14nm vs 16nm for these new cards. I want to clear up some MAJOR misunderstandings for the silicon transistor feature size.
First off, NEITHER of these GPUs will feature TRUE 14nm or 16nm transistors. Intel is THE ONLY company that has true 14nm feature size FETs. (
source_1,
source_2). These FETs are actually 20nm at feature size, but work in the 14nm/16nm description. Globalfoundries (creating the 14nm AMD Polaris) and TSMC (creating the 16nm Nvidia Pascal) can use these terms because their FETs perform roughly at 14nm quality but have a slightly larger footprint.
What does this mean to the users? Its marketing play all over again. All these GPUs should be considered 20nm FIN-FET. Since they are FIN-FETs rather than planar FETs, there are major benefits. The first noticable benefit is power. Reduced size, and using FIN-FET decreases voltage threshold, and decreases current leakage (means major power saving). Size reduction brings in more transistors, and its easier to place transistors where you need them. Since power is reduced, speed can be increased (as seen by the OCed 1080).
All in all, the new silicon shrink for GPUs have been needed for a long time. The benefits will be there. However, don't get confused by marketing! Its not true 14nm/16nm its more close to 20nm