SemiAccurate today has exposed a number of Next Generation NVIDIA Graphics card release plan for the future, the future will be straight across the 20nm and 16nm.

This year the mainstream NVIDIA flagship core Kepler architecture GK104 will be replaced by GM204 the second generation of Maxwell architecture. Now GM104 is the first generation of Maxwell which will be retiring soon.

NVIDIA completely skips 20nm, 16nm four SKUs are Coming

A big GK110 core successor will be GM200.
The GM204 graphics core will have at least four models, expected to be named as:

  • GeForce GTX 880 Ti
  • GeForce GTX 880
  • GeForce GTX 870
  • GeForce GTX 860

All four are expected to release third quarter of this 2014.

In all four models difference is in number of stream processors, frequency, memory is different, there may be up to 15-20 streaming multiprocessor units, namely 1920-2560 CUDAs. Efficiency Maxwell architecture, energy efficiency will be on a large scale and therefore GM204 may not exceed GK110.

 NVIDIA to skip 20nm process, third Generation of Maxwell to use 16nm

The most striking is, NVIDIA 20nm process will be completely ignored. Before and after the 2015 first quarter interim, NVIDIA will launch the second wave of new products, or third-generation Maxwell, the core architecture or GM204, but the process will be used directly 16nm – TSMC plans early in 2015 production of this process, that NVIDIA will be the first time to follow up.
In other words, you can say 20nm GM104 A stepping, 16nm GM204 is a B stepping.

The16nm version might be a GeForce 900 series, of course, possible name change (to be honest) or add a suffix to existing models (if you remember 9800GTX +).
As for now, why they are going to abandon 20nm, there is no clear view, the process may be in one or more aspects of performance, power, area, cost, and may be other people on organization are not very satisfied with the upgrade.