Great exchange guys, I appreciate the insight. What would you recommend as a good starting point for overclocking guides? I did some searching myself and found plenty of guides for the 3930k. Ideally I'd like to stay under 1.375 on the vCore (if possible) using offset voltage, but was also wondering what kind of c-states / LLC settings to use. I'm not looking to break any OC records, but a 4.3-4.5 ghz clock would make me happy!
Guides and threads found so far:
http://www.overclock.net/t/1189242/...anations-and-support-for-all-x79-overclockers
http://forums.anandtech.com/showthread.php?t=2323747
http://rog.asus.com/forum/showthread.php?17627-Rampage-IV-Extreme-3930k-Overclocking-help
I was drawn to your response and your questions.
The OC guide for Sandy E which I found a couple weeks ago -- I haven't been able to relocated with any certainty. You might try Benchmark Reviews or "Benchmark HQ". Oddly, I remember certain things from what seemed to be a very professional and detailed guide.
Since I only have the Sandy K processors, I still found the E guides extremely relevant and more so for X79 boards by the manufacturer of my Z68's -- ASUS. In other words, I can gather wisdom from a lot of guides:
1st priority: Guides which focus on your particular processor and your particular motherboard
2nd priority (not to be excluded): Guides which focus on other motherboards with the same chipset and your processor
3rd priority and not to be excluded: Guides for successor chipsets, possibly same board maker, and later processors for the same socket and successor chipsets. For instance and for me, I've consulted guides meant both for SB and IB K chips and Z77 chipsets.
Immediately, in answer to your post and based on experiences I'd had over the last year, I will tell you to turn off or disable the C3/C6 Report items in the BIOS, but leave the EIST and C1E enabled. I had been plagued by IDLE instability which occurred over 10-day to 30-day intervals, making it very hard to replicate and troubleshoot. After enough "research" of "guides" and causes for BSOD stop-code 124, it is now absolutely clear to me that C3/C6 Enabled is the major culprit for these occasional idle crashes.
What is the drawback for turning them off? Either nothing or next to nothing. Turning them off makes it difficult or impossible to use certain sleep states related only "Hybrid Sleep" settings in Windows. Also, disabling them may result in a slight increase in power consumption in the order of perhaps a couple watts at idle. But if the computer and its overclock have been thoroughly tested for stability, you can still put the computer to sleep and wake it; you can still allow it to hibernate and wake from hibernation.
On the positive side, turning off C3/C6 stabilizes idle voltage. The unfortunate interaction between LLC and VCORE Offset voltage can cause voltage fluctuation in the idle state, when the processor is showing close to or below 1.0V under the EIST/C1E Enabled setting. If it drops too low at one point in time or another, it will generate the 124 BSOD. So you will suddenly notice the disappearance of voltage fluctuation between 0.8 and 1.0, with a stable voltage much closer to 0.98 to 1.0+.
On the matter of LLC, the SB-E guide recommended what I'd gleaned independently from working with my 2600K and 2700K: Do not exceed an LLC setting of "High" (for an ASUS board Z68 or X79). On my boards, the selections are "Auto" (none?), "Regular," "Medium", "High", "Ultra High" and "Extreme." The High setting still allows a vDroop in loaded voltage, which under the more severe stress-tests might be approximately 30 mV. You do not want to eliminate vDroop altogether. Droop is good.
Since you have the Rampage IV board (ASUS), I'd bet that all of these suggestions apply well to your system.
The other recommendations I saw in the SB-E guide were similar to what I might find in an SB-K guide.
Avoid using the "Extreme" duty-cycle and phase-power settings if you can, or unless you've made special attention to cooling VRMs on the motherboard. If you have to, I'd personally found that the duty cycle can be set to "Asus Optimized" which I think is default, while the phase-power setting needed to be set to "Extreme" for balancing current in the VRMs as opposed to temperature balance.
Also, the matter of PLL Overvoltage, which is not the same as PLL Voltage. Avoid using it -- that is, leave it disabled -- unless you're trying for clock settings of 4.7 Ghz or higher. A myth about PLL Overvoltage had been promulgated in Sandy Bridge guides telling users to enabled it beginning with clocks of 4.5. Better off if you can manage your best overclock without it. And the E guide discouraged its use for 24/7 settings, just as it discouraged using "Extreme" on both the duty cycle and phase-power settings.
Also, another myth has been dispelled: It was common practice to disable Spread Spectrum -- a practice going back to Pentium 4 days. You might try disabling it while tuning your system for your clock target and verifying stability. But in the end --
enable it, perhaps do another stability test run to see if other tweaks are needed. But it will also stabilize voltage at both idle and load if enabled.
As for what guides to read specifically, I don't have specific links for you. Just keep searching, and get an idea about which guides are written by people who know what they're doing.
And also, I think you are spot-on for choosing a self-imposed voltage target of 1.375V. That's what I've chosen to do -- more or less.
The only thing to split hairs about with that: there is some consensus that the "safe limit" applies to the extremely loaded, drooped voltage of severe stress-tests. In my case, if the unloaded turbo-speed shows a voltage at just above 1.38, this loaded value would be ~ 1.35V. For that, I think I could still go higher to bring that loaded value up to (perhaps) 1.37. But this is all a matter of choice, desired clock settings, temperatures, etc. If you can achieve your clock target with less voltage after tuning it for consistent GFLOPS in certain tests, why push it higher?
Now with all that, I'm still thinking of bumping up my 2700K to 4.8 Ghz, because it won't much push the temperatures into an uncomfortable range. But I might see unloaded turbo voltage show up in the HWMonitor "Maximum" column at 1.40+. I'll think about it some more, while I continue to enjoy my gaming at 4.7.
Oh. ONe more thing I remember from that guide for SB-E overclocking. Set your "current capability" at 120% and no higher than that. Let me emphasize -- no higher than 120%.
Facebook
Twitter