German 3DCenter.com source reports that expected soon NVIDIA solutions based on GPU GM200 will be deprived of a powerful unit FPU calculations performed on 64-bit operations. Note, the first based on GM200 (it's big brother in the generation of a large Maxwell) was expected to yield solutions in the line of NVIDIA Quadro. Orientation to accelerators for professional use makes NVIDIA create graphics architecture that can efficiently perform non-graphical calculations.
For lack of market gaming graphics unit FP64 (double-precision calculations) will not have any noticeable effect. The main emphasis in the processing of game graphics is calculation with single-precision or FP32. For older video cards on the Big Maxwell it can be even better. Without volumetric unit FPU FP64 on-chip energy efficiency powerful NVIDIA graphics cards can grow very much. But NVIDIA lose market supercomputers where GPGPU-calculations have become the norm.