So... there just seems to be no way to use an GTX 970 properly for mining with Win 10 (1703).

after googling and testing for an entire day...

382.33 + CUDA 8.0: 3 MH/s
372.54 + CUDA 6.5: 9.5 MH/s (with P0)
347.52 + CUDA 6.5: 16.5 MH/s (driver won't allow P0 setting)