And here we go...
For many months now my old system has been rock stable, hell I've even played and completed Dragon Age Inquisition. I then installed the 353.06 drivers and was getting " driver had stopped responding " in Windows, then in the 353.12. This turned out to be a Nvidia wide issue which was then later fixed in the 353.38.
I got a new system, everything was working fine until in my OCD ways of having the latest drivers, went for the leaked ones 353.51. I was getting grey screen of death in Arma 3 and Insurgency. Thinking it was the drivers I downgraded to the 353.49, same thing issue but more often.
I dropped my core down by 50mhz and my memory by 500, it was still happening. Hell I even tried the original bios for laughs, but then I noticed something.
In Arma 3 and Insurgency my core clocks and volts were dropping down to a lower power mode. Which of course I've not seen before as I've always been playing high gpu required games aka The Witcher 3, Dragon Age Inq. In those games the core is always maxed out as are the volts.
Every time this grey screen of death was happening, was directly after the core and volts were dropping to a lower power state, then coming back up again but not in sequence. Sometimes the core would go up to 1480mhz but the volts would stay at 1.1. This is also directly related to the timing issue and power mode that happened with people with Chrome + HW in the 353.38 drivers and why it mainly affects the superclocked even factory overclocked cards.
I downgraded to the 350.12 and noticed the core clocks and volts were far more stable, so they were not going into low power mode as often, also to note many many users on the geforce forums recommend these for fixing grey screen of death and TDR's.
Now I'm using KBOOST which keeps my core/mem/volts maxed out as they should be in games with no drops. ( This is NOT a thermal throttle issue )
Played Arma 3/BP for an hour, Insurgency for the same.