I'll post just to avoid another trivia thread about my puzzlement in troubleshooting.
I had 2x GTX 970 in SLI in one of my systems from Jan 2014 to late last summer or early fall 2016. I began to experience shutdown and startup delays, and thought I'd traced it to my KIS AV/firewall program. At some point, I decided to remove one of the graphics cards, if only for saving power because the system was no longer primary to my gaming inclinations. It became exclusively a system for HDTV and serious business work.
Since then, it seemed as though drivers were behaving in a flakey way. Every so often, the system would be locked in a perpetual "shutdown" screen with the rolling donut. There was an additional complication even though I'd eliminated AfterBurner overclocking and one of the cards: The system feeds a desktop monitor (primary) and my HDTV.
More recently, even for waiting for the delayed "shutdown" screen to reboot (and it would do that maybe half the time), I would just as equally get diagnostic BlueScreenView interpretations of Bugcheck mini-dumps: "DRIVER_POWER_STATE_FAILURE" associated with the Nvidia driver, the Win kernel and TCP-IP.
I finally swapped the 970 cards, loaded up a clean install of this thread's subject driver (which I'd installed to no improved effect earlier), and everything is wonderful again. I could still test to see if the offending card was seated properly in the #1 PCIE-x16 slot -- that's always a possibility.
What a waste of a good part of two days getting this sorted out . . . I even replaced the PSU Seasonic "G Series" 650W with a Seasonic "Gold" X-series 650 I had waiting for another PSU swap in a different system.
But that wasn't the problem, after all . . .