For readers: We were discussing that SB Static Over Clocking has issues with BSOD's at idle, specifically watching online video and there really seems to be no specific fix even if you increase the off set voltage. Perhaps some of you have found a cure.
Not quite. I didn't notice that "static overclocking" was the topic, and I never began my over-clock calibrations in this way: EIST, C1E and other C states were fully implemented.
The "BSODs" and resets at idle EIST are noted in a rash of forum posts including "just plain Windows." that is, "Event ID 41" is the only event viewer trace of some sort of idle instability, but it had occurred for many mainstreamers using OEMs and laptops with the power-saving enabled.
The frequency of the problem cited for these as well as some "SB-K overclockers" was much more troublesome than my case: it would occur one or more times daily, or every couple days at the least. For me, it happens between 10 days or longer -- sometimes a month. I won't trouble to print my spreadsheet log from Event Viewer with days-between-dates. Just assume that I've tracked it as a statistical series.
Possible troubleshooting priorities:
C1E and other C states enabled [but in my case, didn't fix anything by changing to "disabled."
Hardware failure -- RAM, VGA, PSU, etc. At one time, since I noticed my UPS beeped a switchover to battery some few times and a less than a day before my own reset events, I thought the UPS might be suspect, but all these items get a clean bill of health.
After you rule those things out:
OS corruption or buggy drivers -- particularly VGA, IRST, etc.
But, carefully ruling out those factors, eventually one runs across discussions with SB-K overclockers experiencing the difficulty at idle EIST. The sources of instability are cited as LLC in combination with vOffset, with mention of PLL voltage (not the "overvoltage" addressing the VRM control loop) and seemingly "safe" VCCIO settings even below 1.2V. But the observations about LLC and vOffset are the same things for which the 2007 December Anandtech article on the QX9650 cautioned with helpful graphic aids:
http://www.anandtech.com/show/2404/5
. . . and the page that follows.
For SB-K, the caution is the same but less general: "Be careful venturing into "(-) negative offset territory."
None of these basic relationships would have changed much over the span from SB-K through IB- and Haswell-K, nor would the potential for idle crashes -- either with fixed-vcore OC'ing or under the power-saving features. Perhaps some refinements that attenuate the problem were added with IB or Haswell, but I wouldn't know firsthand. I'd have to see the BIOS options for a Z77 or Z87 motherboard to be sure.
Nor were the discussions I found indicative of anything other than "Offset" overclocking with EIST enabled.
I've used some of the features your settings show -- for instance, "Current capability 140%," although I limited mine to 120%, because I intuitively sensed upping the current would accelerate the same chip degradation over shorter time than would just occur naturally -- it happens whether you OC or not. I've found that for 4.6 and 4.7, it's unnecessary. As for LLC "Ultra High" setting, at higher clocks (and per the old cited white-paper) you then are likely to exceed VID at high clocks -- which is "not good."
Some early wisdom suggested using these features for better stability at load conditions, bench or more extreme overclocks. "PLL Overvoltage[Enabled]" was also given as a panacea above 4.5Ghz, but I've since discovered it isn't necessary. While it isn't a voltage per se but merely an adjustment of the control loop for voltage signal, it still has an impact on VRM components, or so I would think -- since setting it to "auto" or "disabled" seems to reduce my temperatures a bit -- beyond simply the perfunctory reduction of PLL voltage itself.
I'd never had a problem with "awake from sleep" with PLL_O____ Enabled, yet many people do. Even so, for this computer, I don't let it sleep as a practice as I do for the other household machines.
But the problem described in various discussion occurs with greater frequency. If a laptop user saw his system reset once every couple weeks of 24/7 usage, he might think it were heat or cosmic rays, but it wouldn't seem initially "serious." A laptop user isn't going to leave his machine running sleep-disabled 24/7/365 with a constant cable-TV feed, or two monitors. In my case, it seems to be extremely borderline and less frequent, but it has a cause.
Speaking of heat and cosmic rays, someone also noted that the occasional nature of my instability may be a result of over-stressing VRM components -- for instance, with settings I already mentioned. At this point, I can only cite small but noticeable temperature improvements at the same clocks and voltages.
I rather doubt that when some here refer to "24/7" it means they leave their computer "sleep-disabled" continuously, running 24/7 video feeds while gaming and other activities simultaneously. Maybe some do. But in my case, I'm running a mult-monitor setup (HDTV and desk monitor) with different resolutions -- again provoking situations raising discussion on various forums with "Event 41" log traces. This then leads to the VGA hardware and driver issues -- already mentioned and excluded -- if not excluded, then "fixed."
Like I said in another lengthy post, this little glitch requires me to watch for it over a month's time. If it doesn't occur -- well, then, it will be "fixed."
But it's not likely a result of my basic over-clocking choices -- not too different from your own, except for trading off "offset" choices against "extra voltage for Turbo." I still don't understand why folks don't avail themselves of this latter voltage feature, because it is "on demand" while offset is effective across the voltage range -- idle or load -- barring the unforeseen idle effects of high LLC settings.
The problem I have is infrequent and totally undetectable during IBT and LinX stress tests. I might even guess that with your own clock settings paired with my usage and peripherals, you might run into it, or you wouldn't without duplicating those conditions. I could run LinX for two days, and it wouldn't arise. But I'd be aging my CPU faster than I'd want to.
The original topic here was "Reason to upgrade from i7-2600K." As far as I can see, it's still "future-proof" for the moment. But if you overclock, you will do troubleshooting. Troubleshooting takes time. And if you want to reap the advantages, you either choose to ignore a minor problem -- or do everything you can to fix it.
If time is money and both are scarce, OC'ing would seem to take more time than it's worth -- if pursuing "last grains of rice." But the choice to do it separates "Enthusiast" and "hobbyist" from the "Mainstreamer" and "Dabbler."