when to adjust voltage when overclocking core i7 2600k?

[EskLuxor]

i have an asrock z68 extreme 4 gen 3 and overclocking the i7 seems fairly simple adjusting the multiplier. i currently have it o/c @ 4ghz. its cooled by a cooler master 212+. if i want to o/c higher will i need to change the voltage?

 

[MrK6]

i have an asrock z68 extreme 4 gen 3 and overclocking the i7 seems fairly simple adjusting the multiplier. i currently have it o/c @ 4ghz. its cooled by a cooler master 212+. if i want to o/c higher will i need to change the voltage?

Eventually. Keep cranking it until it crashes.

 

[Puppies04]

Eventually. Keep cranking it until it crashes.

While strictly speaking this is correct you need to keep a few things in mind OP. Unless you set a manual voltage (which I assume you haven't done) the CPU will request and recieve more and more voltage the higher you push the overclock. Most people find that they can actually force the MOBO to supply less voltage than the CPU thinks it needs at around 4ghz. This is achieved by using "offset mode" in the bios and dialing in a negative "-" voltage adjustment.

First things first though.

Before you do anything else you need to get a couple of pieces of software to monitor heat and voltage. I suggest coretemp or CPUIDHW monitor (preferably both) to monitor heat and CPU-Z to monitor voltage although CPUIDHW monitor does this also. Then grab intel burn test and while the other programs are running perform a "very high" test set to 20 runs and post back here with the max temps you see on each core. Once we have that data we can discuss if you want to OC further and the best ways to achieve this, also if you decide to stick at 4ghz we can discuss ways of reducing the voltage slightly to keep the cpu a little cooler.

 

[EskLuxor]

after i overclocked it i ran prime 95 for 3 hrs. 8 instances of the test poped up assuming i have 8 threads or logical cores due to the hyperthreading. i monitored the temp using REALTEMP. none of the cores went higher then 67 cel. ill grab the software u suggested and report back once the test has been completed.

 

[Diogenes2]

What is your CPU voltage value at 100% load ?

If you are leaving the voltage on ' auto ', chances are it is going up higher than it needs to be for a stable OC..

You should set it manually to establish lowest stable setting, then use the ' Offset " method to dial it in to your stable value.. This will result in the lowest stable setting at all clocks: particulary when the CPU is idle..

 

[cytoSiN]

What is your CPU voltage value at 100% load ?

If you are leaving the voltage on ' auto ', chances are it is going up higher than it needs to be for a stable OC..

You should set it manually to establish lowest stable setting, then use the ' Offset " method to dial it in to your stable value.. This will result in the lowest stable setting at all clocks: particulary when the CPU is idle..

^ This. Offset mode (as opposed to full manual) will also let you use speedstep and c1/c3 effectively to minimize power consumption and, more importantly, heat!

Also, in addition to offset mode, load line calibration (LLC) is a useful BIOS setting that will help you avoid vdroop and even undervolt at your chosen overclock.

I'm running my 2500k at 4.5 ghz with a -0.04v offset and LLC at 50% (in my BIOS, it goes in 25% steps from 0 to 100%).

At idle, I'm supplying 0.976v (VCore) while the CPU is asking for 1.017v (VID -- RealTemp will tell you your VID if you click the running stopwatch button in the upper right). It's rock stable at idle despite the slight undervolt. Temps around 25-30c idle depending on ambient temp.

Under load, I'm supplying 1.304v while the CPU is asking for 1.346 (and sometimes 1.370). It's rock stable there as well, with temps around 45c under normal load, and around 65c under insane "why would you do this to a CPU" load. Without LLC at 50%, however, I was seeing substantial vdroop (down to around 1.276, creating a difference of 0.07v to 0.1v between VCore and VID) that was causing BSODs under the aforementioned insane load.

Looking at those numbers, you'd think I could just remove the offset (keep offset mode on, but switch the offset to "auto" or 0.00v). But leaving it on lets me run a significant overclock (3.3 to 4.5) despite undervolting at idle and load, which will, in theory, increase the longevity of the CPU and keep temps as low as possible.

Moral of the story: Offset mode + LLC should get you well on your way to a stable overclock with the lowest possible voltages.

 

[Puppies04]

after i overclocked it i ran prime 95 for 3 hrs. 8 instances of the test poped up assuming i have 8 threads or logical cores due to the hyperthreading. i monitored the temp using REALTEMP. none of the cores went higher then 67 cel. ill grab the software u suggested and report back once the test has been completed.

Will keep an eye out for your reply.

 

[EskLuxor]

ok i installed all applications, run both cpu monitoring programs and intel burn test. set program to very high and 20 runs as suggested. core temp lists max temps as follows, core #0 59 core #1 63 core #2 65 core #3 61. it executed the test for 358.94 seconds and a critical error reared itself. said yr system was found to be unstable under intelburntest. check cooling/lower o/c etc, etc.

 

[jimpatrick]

ok i installed all applications, run both cpu monitoring programs and intel burn test. set program to very high and 20 runs as suggested. core temp lists max temps as follows, core #0 59 core #1 63 core #2 65 core #3 61. it executed the test for 358.94 seconds and a critical error reared itself. said yr system was found to be unstable under intelburntest. check cooling/lower o/c etc, etc.

that would be @4ghz right?first time running ibt or where you able to finish some runs before?

 

[EskLuxor]

correct @ 4ghz and it was the first time running intel burn test.

 

[Puppies04]

ok i installed all applications, run both cpu monitoring programs and intel burn test. set program to very high and 20 runs as suggested. core temp lists max temps as follows, core #0 59 core #1 63 core #2 65 core #3 61. it executed the test for 358.94 seconds and a critical error reared itself. said yr system was found to be unstable under intelburntest. check cooling/lower o/c etc, etc.

Did you make a note of the voltage reported in CPU-Z while the test was running?

Your temps look good but that warning indicates you might need a little more voltage, post back with the numbers and we will see.

 

[EskLuxor]

running again right now, am i looking for voltages in cduid hardware monitor or in core temp?

 

[EskLuxor]

ran again, success this time. test ran 3612.03 seconds max temps reported by coretemp were core #0 60 core #1 64 core #2 67 core #3 63. max cpu vcore is 1.28

 

[cytoSiN]

running again right now, am i looking for voltages in cduid hardware monitor or in core temp?

In theory the goal is to match the VCore (actual voltage) to the VID (voltage the CPU wants). You can see VID by running Realtemp and clicking on the running stopwatch button in the upper right to change the display to VID (there are other ways, but that's an easy one). Obviously CPU-Z and just about every other program shows VCore. The closer those two numbers are, the more stable your overclock, at least in terms of voltage. As I mentioned in my post above, it's often possible to undervolt and maintain stability with the right combination of voltage offset and LLC.

ran again, success this time. test ran 3612.03 seconds max temps reported by coretemp were core #0 60 core #1 64 core #2 67 core #3 63. max cpu vcore is 1.28

Congrats on the second stable pass! Those are pretty good temps for 1.28 with IBT too. Mine go up to 70, and sometimes slightly over, running IBT with those settings at 4.5GHz (and about 1.3 VCore, so similar to your voltage).

 

[Puppies04]

ran again, success this time. test ran 3612.03 seconds max temps reported by coretemp were core #0 60 core #1 64 core #2 67 core #3 63. max cpu vcore is 1.28

Seems good, be aware that these chips can't require a little bit more voltage after they have been run for a few weeks so if it becomes unstable again just tweak it slightly.

 

[EskLuxor]

thanks for the reply's and information.
any suggestion why the test might fail first and w/o changes pass the second time?
my vcore supplied by cpuid hw monitor .96-1.0v while realtemp reports the vid as 1.2360. (idle)
should there be that much of a variable? cytosin, you said from a voltage standpoint the overclock prolly isnt very stable. i guess i should have mentioned that im using asrock extreme tuner utility, an inside windows o/c program, should i go old school and use the bios instead? i believe i disabled speedstep in the bios, was that a good idea?
Realtemp shows core frequency dip between 2922-3991mhz. i thought if i disabled speedstep that the overclock would not downclock?

 

[cytoSiN]

thanks for the reply's and information.
any suggestion why the test might fail first and w/o changes pass the second time?
my vcore supplied by cpuid hw monitor .96-1.0v while realtemp reports the vid as 1.2360. (idle)
should there be that much of a variable? cytosin, you said from a voltage standpoint the overclock prolly isnt very stable. i guess i should have mentioned that im using asrock extreme tuner utility, an inside windows o/c program, should i go old school and use the bios instead? i believe i disabled speedstep in the bios, was that a good idea?
Realtemp shows core frequency dip between 2922-3991mhz. i thought if i disabled speedstep that the overclock would not downclock?

That's your problem right there. I advise people to NEVER use the auto OC. For one thing, it usually sets voltage to manual, without an offset, which effectively disables voltage stepping (often referred to as c1/c3). What that means is that when Intel speedstep ramps your multiplier down to x16 at idle, your voltage stays constant. This is a waste of electricity and it's bad for your CPU for obvious reasons.

Then again, your VID seems pretty high for idle, but that's because you disabled Intel speedstep. In my opinion, for people who overlock as part of their everyday computer use, there is no reason to disable any clock or voltage stepping. These features are really good at saving power and increasing component longevity. The only reason to turn them off, in my opinion, is to run an insane overclock (custom H2O or LN2, e.g.), where benchmarking/folding/etc. are your only real focus.

In your case, I would go into the BIOS and reset settings to default. Then set the voltage to offset mode, but don't actually enter any offset for now. Just leave it at "auto" or "0.00." We can tweak this later. Then make sure that speedstep AND c1/c3 are ON. Once you've done that, increase your clock to 40, and leave everything else alone. Then boot up, open Realtemp and CPU-Z, and run Prime95 and/or IBT, and report your voltages/temps here. Please post that info at idle and with Prime95 or IBT running. Also, sometimes the VID will have two different "load" values depending on how heavy the load is. As you can see from one of my previous posts, my 2500k usually asks for 1.34v under load, but sometimes, under "heavy" load, it will ask for up to 1.37v. If you notice anything like that, post it for us.

 

[EskLuxor]

ok thanks for the info. i have a busy few days so i might not be able to do this until sunday morning. ill keep u posted, thanks again for helping.

 

[EskLuxor]

ok, ive loaded default in bios, turned cpu core voltage to offset mode. one thing about it is there is no "0.00" or "auto" its only "-0.005" or "+0.005" intel speedstep in enabled again. should turbo boost be enabled? now my pc wont boot

 

[cytoSiN]

ok, ive loaded default in bios, turned cpu core voltage to offset mode. one thing about it is there is no "0.00" or "auto" its only "-0.005" or "+0.005" intel speedstep in enabled again. should turbo boost be enabled? now my pc wont boot

Yes, enable Turbo. As for offset, you should be able to set it to 0. If not, try +.005 first. If you can't to get it to boot with the 40x multiplier and offset mode, set voltage to auto and see if that helps. We can come back to tweaking it once you're booting.

 

[EskLuxor]

ive pushed the clear cmos button on the motherboard. still wont load

 

[Puppies04]

ive pushed the clear cmos button on the motherboard. still wont load

Have you checked your bios to make sure it is trying to load the OS from the correct hard drive since you reset it.

 

[cytoSiN]

Have you checked your bios to make sure it is trying to load the OS from the correct hard drive since you reset it.

Good point.

Also, look at this for a list of settings for your motherboard (with a 2500k at 48x, but everything else should apply for helping you find a stable boot).

Also, we haven't discussed ram, but just to be sure that's not what's holding up your boot, make sure you set the appropriate XMP profile after resetting BIOS settings. Sorry for not mentioning this earlier.

 

[EskLuxor]

sorry no luck keep getting "windows failed to start" then i select start windows normally, then BSOD. no luck on a restore point or trying to repair errors either.

 

[EskLuxor]

reset the xmp profile, ran a windows memory test from install disc, check the list from overclock.net, checked if it was booting from primary hd. i think im kinda outta options at this point.