2500K Occasional POST Failure

AXELfrieman

Junior Member
Aug 17, 2015
5
0
0
Hello everyone, I'm new in town on my whirlwind journey to fix my OC. Hopefully this will be my last stop! A few weeks ago I decided to finally overclock my aging 2500k due to it finally becoming the bottleneck in my system at stock speeds. I've not overclocked a system since I juiced up my old Core2Duo to play a Crysis mod many moons ago. Needless to say the Sandy Bridge experience was quite different than I remember. After setting the multiplier to 45 and letting SVID handle the vcore (1.3811) all seemed well. However I am now noticing an occasional failure to POST. While running the OC is stable as I have stress tested it and it has seen real-world testing of rendering for over 12 hours. In an attempt to resolve the POST failure I have tried to raise the vcore to 1.40 but this does not effect the voltage. It seems that SVID is overriding my entry. Help me, AnandTech. You're my only hope.

Components:
2500k
Mobo. Z68X-UD3H-B3
16GB RAM@1600 (BIOS is reading it as 1300 though, haven't tackled this yet)
H60 cooling
 

Schmide

Diamond Member
Mar 7, 2002
5,581
712
126
Increase your ring bus voltage. Don't go straight to 45, work your way up there.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
"Work your way up."

That's the point of it.

I'm not as eager to provide advice about a Gigabyte Z68 board. My Z68's are ASUS P8Z68-V Pros.

I looked at the (non-UEFI) BIOS screens. What someone called "ring voltage" I'm hoping was meant as something else.

I'm only guessing that Offset voltage falls under "Dynamic VCORE." And I don't see a secondary voltage adjustment for "Extra voltage for Turbo."

Otherwise, it's a "performance" motherboard and looks solid.

At 4.5Ghz, you shouldn't need to play with "QPT/VTT" or VCCIO, but the symptom the OP describes seems to intersect with some descriptions in the Anandtech review for that board model.

But before anything else -- and especially fiddling with that or any voltages -- start by setting the board up at default settings and choosing the XMP profile for RAM @ 1600 and the stock timings. Assure yourself that your RAM settings aren't boosting your bCLK above 100 -- that the RAM is running at 1600 Mhz.

Then, if "Dynamic VCORE" reveals Offset voltage settings, chose "+" and the smallest increment nearest 0, and start working from there. If you even need LLC (and there are some ten notches on that board), you would choose a setting that still leaves 40mV of vDroop.

Also, leave EIST and C1E enabled, but turn off "C3" and "C6" reports or the settings referring to those features. You'll still be able to sleep and hibernate the machine, but less likely to have any unexplainable idle instability. Much less likely.
 

Schmide

Diamond Member
Mar 7, 2002
5,581
712
126
Ok what I was referring to on that board would be the CPU PLL. Basically what I think of as the default power plane for the CPU. If you increase that from 1.8v to 1.85-1.95v it will increase voltage to the what some call the uncore.

That should help on cold boots.

Edit: Here is a better explanation as I may be a bit off.

http://www.overclock.net/t/1189242/sandy-bridge-e-overclocking-guide-walk-through-explanations-and-support-for-all-x79-overclockers

CPU PLL Voltage Override (Overvoltage): What the Heck does it do?
So I asked that question to an Intel Overclocking Engineer his explanation was roughly: We went through the BIOS settings trying to find setting that if changed could help overclock our CPUs further. We came across this setting. Think of the CPU PLL voltage as a voltage that is provided to the CPU, but then “clipped” down to an approximate voltage. No matter what that input is whether 1.3v or 1.9v it is clipped (hypothetically let’s say 800mv after clipping (he didn’t say how much)) that way other devices can use the PLL voltage and clip to what they need. The CPU PLL Overvoltage allows for less clipping of that voltage. It can also reduce the lifespan of the CPU, but nothing noticeable.
So those of you who think that increasing your PLL voltage will help with that setting, it really doesn’t. But with SBe I have found that increased CPU PLL can help stabilize higher frequency overclocks. That wasn’t the case with SB.
 
Last edited:

Charlie98

Diamond Member
Nov 6, 2011
6,292
62
91
Are you sure you need all that voltage? I've got my 2500K OC'd to 4.1 on a lesser Gigabyte mobo and it's undervolted. Maybe try it with stock voltage?

And I agree with Schmide... maybe work your way up from 4.0-4.1GHz...?
 

AXELfrieman

Junior Member
Aug 17, 2015
5
0
0
Bonzai- I'll try setting the RAM to 1600 again. I had received a memory crash at one point and assumed it was RAM related, but it also could have been a generic instability crash. I'll also enable C1E and EIST again, I had disabled all three as usual. I see now that Gigabyte was pretty stoked about their dynamic vcore when it was released, so this may be the only way for me to interact with my voltages due to some reading I've done on dynamic vcore. I'll give the DVID method of OC a try.

Schmide- From what I've read of the PLL is that I could lower it to possibly reduce my CPU temps. However if I understand the math, by reducing my internal phase lock loop (PLL) then I would have to boost my BLCK to achieve the same overclock frequency. Stock is 1.8v, and Intel states 1.89v maximum. Not a lot of leeway. I'm not sure my kung fu is good enough for this yet. :'(

Charlie98- I've not added any voltage, as the voltage supplied has been done automatically by the Serial VID, which is supposed to scale to the appropriate amount of power needed at a certain frequency multiplier. When changing the vcore manually the SVID feature should be deactivated and the vcore input taken instead. This however is not occurring and all of my attempts to directly change the vcore are ineffective.

Going reset the mobo, lock in the RAM, and attempt the DVID method of overclock then will report back with the findings. Thanks for the assistance everyone. :D
 
Last edited:

Charlie98

Diamond Member
Nov 6, 2011
6,292
62
91
Gotcha... I don't have that problem with my board. I will recommend, however, you leave the BCLK at 100, Sandy and up don't like higher. When I first tried to OC my board, I used the Gigabyte Smart 6 OC utility... which relies on bumping up the multiplier and the BCLK... and I had constant crashes. Once I went in and started setting things manually, leaving the BCLK alone, everything was fine.
 

Schmide

Diamond Member
Mar 7, 2002
5,581
712
126
Schmide- From what I've read of the PLL is that I could lower it to possibly reduce my CPU temps. However if I understand the math, by reducing my internal phase lock loop (PLL) then I would have to boost my BLCK to achieve the same overclock frequency. Stock is 1.8v, and Intel states 1.89v maximum. Not a lot of leeway. I'm not sure my kung fu is good enough for this yet. :'(
Thanks for the assistance everyone. :D

If you're getting cold boot problems. Temps are not the cause of it.

It would be good to report your temps from 4.1-4.5ghz if you are serious about getting this chip there.

Also do some OCCT or like testing before going to games.
 

AXELfrieman

Junior Member
Aug 17, 2015
5
0
0
Here is a pic of some monitoring while I was running Prime95 this morning. The list of changes:
  • Cleared CMOS
  • Updated BIOS ver.
  • Activated XMP to resolve memory speed (very embarassing)
  • EIST, C1E, C3/C6 left on Auto
  • Vcore set to Normal
  • DVID enabled at +0.000v
KGxBsLm.png


Since making these changes the system has booted properly so far. I'll be restarting it some more throughout the day to keep testing it.

P.S. - Which voltage in the image above is the most accurate? I've read the VID in Core Temp is inaccurate and that CPU Vcore under the Mobo. for HWMontior is the correct reading. The CPU Vcore coincides with the Core Voltage within CPU-Z and I am therefore more apt to trust the data.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
Bonzai- I'll try setting the RAM to 1600 again. I had received a memory crash at one point and assumed it was RAM related, but it also could have been a generic instability crash. I'll also enable C1E and EIST again, I had disabled all three as usual. I see now that Gigabyte was pretty stoked about their dynamic vcore when it was released, so this may be the only way for me to interact with my voltages due to some reading I've done on dynamic vcore. I'll give the DVID method of OC a try.

Schmide- From what I've read of the PLL is that I could lower it to possibly reduce my CPU temps. However if I understand the math, by reducing my internal phase lock loop (PLL) then I would have to boost my BLCK to achieve the same overclock frequency. Stock is 1.8v, and Intel states 1.89v maximum. Not a lot of leeway. I'm not sure my kung fu is good enough for this yet. :'(

Charlie98- I've not added any voltage, as the voltage supplied has been done automatically by the Serial VID, which is supposed to scale to the appropriate amount of power needed at a certain frequency multiplier. When changing the vcore manually the SVID feature should be deactivated and the vcore input taken instead. This however is not occurring and all of my attempts to directly change the vcore are ineffective.

Going reset the mobo, lock in the RAM, and attempt the DVID method of overclock then will report back with the findings. Thanks for the assistance everyone. :D

Again, there could be differences between Z68 boards of different manufacture. But there has long been a general consensus that you should be able to get reasonably high clocks with a PLL Voltage set to ~1.70V.

You shouldn't need PLL-Overvoltage enabled unless you're trying for 4.7 Ghz or higher, and on the right chip (and I seem to have two as 2600K and 2700K respectively) 4.7 Ghz doesn't require enabling PLL Overvoltage.

VID vs VCORE and the Monitored values: I don't see anything out of whack in any of the three monitoring programs. CPUZ reports the VCORE; both VCORE and VID are reported in HWMonitor, and the CoreTemp value of VID is consistent. Nor should I be irritated here that you're using a fixed voltage approach to this, but it will easily transition to enabled EIST, C1E and "offset/dynamic" mode.

Also, there should be a newer version of HWMonitor (such as the one I'm using) which reports more information.

Again -- There is absolutely no reasonable need to leave the C3/C6 Report items of power-saving enabled. I say this because you may even experience idle-level instability -- often so infrequent that it defies easy troubleshooting -- with those items enabled. After that, make sure you disable "hybrid sleep" in Windows, but you can use basic Sleep and Hibernate. With the C3/C6 Disabled, idle power consumption may be a few Watts higher: it could make the difference between 4W and 10W. You've already done enough for the "Green" cause without the drawbacks of C3/C6, so . . .

Further, you might try adjusting and testing your overclock settings with Spread Spectrum disabled, but you should be able to Enable Spread Spectrum, re-test -- and then find that both idle and load voltage variation has been slightly attenuated -- a good thing.

Again -- I looked through the BIOS screens for that board, and I think they were shown in an Anandtech review. I couldn't find "Extra Voltage for Turbo," but do another look-see to find it. With both an vOffset setting and this "Extra V for turbo," you should be able keep a lower positive offset value and address your turbo clock stability with the latter.

Folks who overclock with Offset alone -- maybe because they have to do so -- need to "mind" their LLC setting, just to assure that idle voltage doesn't drop too low. I'd say if idle voltage doesn't vary too much below 1.000V, you're "good to go." That's with EIST and C1E enabled, by the way.
 
Last edited:

AXELfrieman

Junior Member
Aug 17, 2015
5
0
0
Thanks for the help duck, it is already a much more reliable OC than anything I had cobbled together myself. My definition of VID was incorrect leading to the vcore vs VID confusion, I now see that it's all good :cool:. I bumped up my offset by 0.015v due to another POST failure, figured this would be a good increment. In regards to the C3/C6, I wasn't aware of the occasional instability and will disable it and hybrid sleep. Thanks again for all of the help Bonzai, in the future I may just stick to the EVGA/ASUS mobos.
 

AXELfrieman

Junior Member
Aug 17, 2015
5
0
0
A short archival update: the failure to post seems to be related to the Gigabyte z68 boot loop issue. If experiencing similar problems I suggest replacing your board with that of another manufacturer.