Bumping this thread for good measure. :>
I must be doing something wrong. After finishing up a few lose ends in WoW before I started overclocking at all, I sort of bumped into a few brick walls. First of all; BIOS update. Smart as I am I managed to not save my settings so I'm at this again. I sort of think I remember what kind of settings I was having but you never know.
I started with x47 on my 4770k, I believe I managed to hold it steady for a couple of hours before at 1.35 vcore and 1.95 vccin. I try to not let VCCIN go more than +0.6 above vcore, 0.5 preferably. OCCT would throw out a myriad of issues.
My OCCT settings for all my testing have pretty much been the Infinite test type with 85c as the max temp stop. Large Data set, automatic threads and 64bit test version.
It might also be worth noting that my uncore and vring is set to x33 and 1.15v, underclocking it while getting stable, and enabling LLC Level 8, I believe C States are set to automatic by default on Asus boards so I haven't bothered enabling them. XMP is turned off and all voltage settings have been set to fully manual mode. Balanced power settings is enabled in Windows.
During my entire testing period it would BSOD with either 0x101 or 0x124 - both which indicate an increase of vcore is needed/decrease is needed -, even at 2 vccin and 1.39vcore it would still toss out the 0x101 error. WHEA_UNCORRECTABLE_ERROR and WHEA_WATCHDOG_TIMEOUT respectively.
I jump down to x46, set vcore to 1.32 which I've ran Aida64 8h+ with last December, vccin to 1.85 and it's still sporting funny 0x101 errors, I up the vcore by 0.05v and all of a sudden it sports 0x124. I continue to up it by 0.05v and I get the 0x101 error again. I actually got it running for more than 1-2 minutes once, but after 4m:21s it would stop the test because one of my cores reached the max temperature threshold (85c), I've never been able to run it for that long again, besides on stock clocks.
For good measure I completely formatted my system and redid the same tests above with same settings / increments and see the same behavior.
Alternatively, x264 from the Haswell overclock net thread seems to do well. I ran it for 18 minutes -- x46, 1.36 vcore, 1.9 vccin without any errors but quit it as soon as temps started reaching 85c and above.
There must be some black magic with OCCT that I'm missing......