the only people that have burned sandys are the people who were pushing 1.5 uncore voltages, so at peak they were doing around 1.55.
they say 1.45 is also dangerous, but that's really only because with 1.45 you need extra high (level 4) or extreme (level 5) load line calibration.
--this puts peak voltage around 1.5 -- 1.55 again. So that seems to be where the problem is coming from.
heat is a none issue below 80C, as long as you don't do any 24/7 stuff..
In my 98F room, my 4.8ghz hits 86C on IBT, 88 on IBT extreme mode
Zubbs and maybe another had noted the low-power-state or idle-voltage problem with choosing negative offsets. My offset is positive, and the EIST idle voltage is always precisely 1.008V -- slightly above the range reported pertaining to unstressed crashing or BSOD. At one point, I think I was just over 0.88V to 0.98 during the OC'ing ordeal.
I'm not an engineer. I do understand industrial or applied statistics, and I have a good idea as to how specs are set by manufacturers after sampling and testing, or how that might be related to warranty policies of the firms.
Users and enthusiasts don't get the testing results and full statistical frequency distributions from the manufacturers; they get "specifications." So you have to just imagine frequency distributions around those specs. That's how we might conclude that a "5% overage above a maximum safe voltage" is not a big deal. But it's also why -- once I obtain stable settings -- I may drop my load speed setting back 50 Mhz, or raise the offset just a notch before calling it quits.
That sort of approach still depends on "specs" -- if any adequate specs are actually published -- which may be meant to assure that any RMA of product over a warranty period would occur for less than 1% (or some small %) of product sold -- ultimately a matter of cost-accounting. Such specs are likely therefore to be very, very conservative.
But one has to choose either an overly-cautious awareness of these things, or lean toward "uncharted territory." Even so, we agree -- we've only seen "CPU-death" from some small number of people who had gone beyond "uncharted territory."
It all boils down to "what you know," "what information is available," and an awareness of "what you don't know -- but only guess." Knowing, deducing, suspecting . . . guessing, and believing are all different things. Belief may be equivalent to "Knowing" in religious matters, but it isn't the same at all in real-world, practical things.