Originally Posted by Partol
How can fluctuating p-states kill a card? That makes no sense to me.
Only ways I can imagine that a video driver could kill a card is by either turning off the fan/fans or allowing too much voltage/current/power.
Originally Posted by Masked
Most of the cards that have reportedly gone down are Fermi as well.
Fermi was a terrible run all-together in terms of heat and clocks. Most of the Fermi run, particularly vanilla, were OC'd already to meet stock clocks.
Add in P-States and the apparent load fluctuations not to mention the voltage hops that are apparently occuring on-top of possible P-state issues.
That's a recipe for a blown Fermi.
You also have to remember that the VAST MAJORITY of common users that blow their cards out, won't know how to take screenshots, don't know about forums and will simply call Alienware or what-have-you for an immediate replacement because they just plain Jane express installed the new drivers and don't want to deal with those issues...That's just how it is.
Back in the day, when I was with AW -- I wrote a mobile driver for a new launch.
Swapped P-states with OCP and also made the mistake of swapping several load profiles. -- Bricked a solid 100-200 customer's cards before we figured out what the cause was...
In the case of a Fermi, most can't hack stock clocks 24/7 which, is actually the core of the issue. You're taking a hot card that already has issues, adding 100% load profiles with a P-state fluctuation -- That's just a dead card.