Yup, 1.36V, it probably needs much less than this but I haven't touched this in the bios because I'm still getting comfortable with the mobo.
Now I have notice this mobo does some real jackass stuff like automatically overvolt my ram to 1.65V even though it is 1.5V through and through (even set in SPD for that).
So who knows, the CPU could be asking for 1.1V and the darn thing is giving it 1.5V for giggles. (its a maximus iv extreme-z, aka mive-z)
If my memory serves, you're running a Maximus IV Extreme-Z? Of course -- you're still "getting comfortable," but already familiar enough to note differences between reported and directly-measured voltages.
Speaking of Sandy Bridge voltages and an over-clocking possibility-set that includes a "fixed vcore" and an "auto" vcore, we're going to see continuing confusion as we incline to continuing the discussion.
With "auto" vcore and "offset" and/or "add'l voltage in Turbo" set, stress-tests are going to show mostly the loaded voltage values. A little more observation and you can see the equivalent of "idle" voltage (with Turbo speed) before or after the stress-test, as the monitoring software continues to log new values. Those loaded voltage values may give some comfort: at about 4.6 Ghz, they should be in the vicinity of 1.30V unless (a) the motherboard's default settings of "auto-everything" are applying unnecessarily higher voltages, or (b) the over-clocker wasn't too keen about stress-testing toward a minimum set of settings.
LLC may throw in a minor factor in determining what those two voltages -- load and idle -- might be. For my build, I use the second-lowest LLC setting and still show a voltage range between 1.30 and 1.35+. There's no doubt to me that allowing the motherboard to just "do its thing" will push those values a tad higher.
Of greater concern to me, not sure how important it is, regular usage even with games won't produce the lower stress-test "drooped" value -- it will be somewhere in between. So as we flirt with voltage settings close to, slightly beyond or above some "safe-voltage" upper-boundary, we should also consider that the stress-test load-values are less likely to occur over normal use.
Again, it's all a matter of probability applied to things that (more or less) follow predictable patterns of degradation over time. AigoMorla's "seat of the pants" ranking of voltage levels and longevity (in warranty-period units) may be good ballpark numbers, and I'm inclined to accept his experience in these matters (per the joke about buying from him a used socket-1366 CPU). There's always the chance that you might find a CPU that can be volted 24/7 @ 1.40V [and would it be "load" or "idle"? . . . ] and see it last two warranty-periods or six years. There's also a chance that it may last less than one period or three years.
So what are those probabilities? Most of us are unlikely to have a clear idea. So we either pick a rule of thumb (1.35, 1.37 . . . whatever), or follow "test-pilot" behavior and "Hang our a**es over the edge . . . . " For the rule-of-thumb, it becomes a matter of whether it applies to "load" -- or "idle." In my rule-of-thumb, I'll incline toward the "idle" number.
