2500K vcore does not drop in idle

I did a working OC a while ago, now I'm a bit rusty on the matter because in the past months I kept my cpu @stock due to high seasonal temperatures.

Now I want to try again some OC on my 2500K. For the sake of completeness, I have a MSI P67 GD65 board and 2 sticks each of 8 GB G. Skill Ripjaws 1600 MHz ram. The cooler is the mighty Noctua NH-D14.

It seems I'm able to reach 4.5 GHz with 1.260 v-core.

The problem is that, even having all power efficiency features enabled (C1E, C-states, Eup 2013, speedstepping), when idling the multiplier drops from 45x to 16x but the vcore does not!

It drops just a little bit from 1.260, something like a mere 1.25 V and nothing more.

While, without OC, with multiplier dropping to 16x, the vcore drops to less than 1 V in total!

Why is this happening?

Among the other OC settings in my BIOS I currently have:

- spread spectrum: off
- v-droop: auto (low v-droop is the only other possible choice, but it dumps a LOT more voltage under load, so I prefer avoiding it)
- xmp enabled

For the CPU OC I'm manually setting the v-core and changing "ratio limit" of each core to 45 from the BIOS.

I've also tried setting turbo boost off and changing the "global" cpu ratio to 45 but nothing changes.

In both cases, another weird thing is that cpu-z reads more than 1.260 v-core during normal usage.

Under stress test, it even jumps to 1.288 V!

Any help is greatly appreciated!

When you set your voltage to manual then set a number value for it, that is the voltage it stays at. Sometimes there is vdroop under load (voltage drop) and sometimes your motherboard will provide more voltage under load. Usually there is a setting called load line calibration that controls that. I have my 6700k overclocked to 4.5 GHz with 1.3v in bios. And I have my llc set to level 7 so my vcore at idle is 1.296 and under load it jumps to 1.312-1.328v. It never drops below 1.296v because it is set to manual. I don't enable c-states but even if I did my clock speed would drop but not my voltage. That is completely normal if you have your vcore set manually.

Pretty much everything you are describing are normal characteristics of overclocking.

So the point is: if I set the v-core value manually, it does not drop regardless of the multiplier and power saving features.

According to this then, if I set v-core on auto with 4.5 GHz, the v-core should drop as the multiplier drops too in idle.

Should I try this and see which v-core value is reached under stress test? Of course it could be a problem because auto could dump much more voltage then actually needed.

What options do you have in bios for overclocking mode? Auto? Manual? Adaptive?

Auto or manual for the most part.

No your vcore will not drop as the core speed drops at idle. That's what I'm saying to you. Not if you set it manually. With adaptive you can set a slightly lower idle vcore than your load vcore but your motherboard is already giving it more under load so find your lowest stable load voltage, find out how much it increases under load and set the vcore around that much lower in bios. Or use adaptive or llc. For me llc is the easiest and most reliable. My vcore at idle is .032v under what it is at load. And with llc I know no matter what my vcore will never drop below 1.296v and never go higher than 1.328. And I idle under 1.3v which is exactly what I want and exactly why I run in at 4.5 Ghz instead of 4.7.

Unfortunately I do not have any adaptive feature nor llc in my BIOS.

I tried something a little different: I've set the multiplier to [45] and left the v-core on [auto].

In this case when idling, the v-core actually drops to around 1 V. Under load, it goes up to 1.320 V, nothing more.

Do you think this is a good solution?

If it's truly stable at both idle and under load it is a good solution.

Do you have Offset? If the board has that option then you can set it so the Vcore will raise and lower with load and multiplier.

1.32v for 4.5 Ghz is absolutely fine if it is stable. I pumped 1.425v into my 2600k for over two years at 4.6 Ghz and it was fine. Those 32nm chips can take a little more voltage. It is recommended to keep those chips under 1.4-1.45 so you are fine. Some people would argue that the constant fluctuation can be hard on the transistors as well but for sure anything under 1.35v on a manual setting is perfectly ok.

wingman99 said:

Do you have Offset? If the board has that option then you can set it so the Vcore will raise and lower with load and multiplier.

Click to expand...

I cannot find it in my BIOS. Is this offset thing also known with some other names?

Should I post a screenshot of the OC setting page maybe?

Rainwater said:

Some people would argue that the constant fluctuation can be hard on the transistors as well but for sure anything under 1.35v on a manual setting is perfectly ok.

Click to expand...

As I said I can only achieve low voltages under 1 V when setting v-core on auto, not manual.

It seems that normal workload and torture test too do not require more than 1.320 V according to the auto v-core setting.

Can my cpu also be stable at 4.5 GHz with a lower v-core? Yes, it can with a 1.290 V v-core and I've proved it by using the manul v-core setting.

But, doing so, I lose the power saving features and the v-core does never drop below 1.250 V.

So, unless I can find this LLC or Offset setting in my board, the question is:

Option A: 4.5 GHz with auto v-core, 1.320 V max under load, 1 V when idling
Option B: 4.5 GHz with 1.260 V manual vcore, 1.290 V max under load, 1.250 V when idling (does not drop below that)

Option B, hands down.

That said, what is the windows power plan set to? If its performance, set it back to balanced and it shoudl drop.

Mine won't drop but that also could be because I use Load Line Calibration. If I set it to balanced or even power saver clock speeds drop but not voltage. But seriously bud, You can run your chip at 1.26v forever and not really see any degradation. I wouldn't stress about your voltage dropping below a volt at idle. Go with the manual setting because long term it is still going to be a better and safer way to go to have your vcore always under 1.3v and only fluctuating up or down .03v than to have it over 1.3v and fluctuation up and down at a much higher gap. Just having it under 1.3 at all times for months and years is a better idea than the alternative.

arkantos91 said:

I cannot find it in my BIOS. Is this offset thing also known with some other names?

Should I post a screenshot of the OC setting page maybe?

Click to expand...

Post a screenshot maybe I can help. With my motherboard I use all the Stock Intel power saving features and I overclocked to 4.5GHz, the Intel features save power and wear. The only thing I changed is Bios is the multiplier to 45 and Dynamic DVID to 0.075.

So the main opinion here is: better having less fluctuation and going for a manual and lower v-core (option B)

Here is how option B in BIOS looks like.

Can you find any LLC or offset feature?

Even with manual v-core set at 1.260 V in BIOS, in actual use cpu-z shows higher values. Mostly 1.280 V but during stress tests it even jumps to 1.3 V

I have spread spectrum off and disabled low v-droop so I don't understand this crazy voltage variance (almost 0.5 V)!!!

Thanks.

That's a .05V difference and not that much.

Spread spectrum has nothing to do with things really..you can leave that enabled of you want.

It says VDROOP control. Since LLC prevents vdroop, I'm betting that is the LLC setting.

VDroop Control has two options:
- auto
- low v-droop

Which should I use?

Well for one man most Cpu's operate at a slightly different voltage than it is set in Bios. Mine does as well. If your operating vCore at idle is 1.28v and under load it is 1.3v that is only .02v increase which is exactly what you want. I don't understand what the issue is. We've all given you suggestions, we've all been doing this for at least a couple years now and we all agree on the same thing. Check it out bro, After about 24 hours total of stress testing with p95 two different versions, Realbench, and Aida 64 I have found that the lowest vCore I can be at and pass p95 version 28 is 1.328v I am not 100% stable at anything lower than that. If I set my llc to level 6 and my vCore to 1.33v I will idle at 1.328v and stay at 1.328v under load. Now flat out hands down now trust me on this HANDS DOWN your i5-2500k Processor is going to require more voltage under load to maintain stability than it is at idle. Period. I am not 100% stable ar 1.296v I need 1.328v So why the **** would I want to idle at 1.328v all the time at moments like these when I'm starting to feel like I'm wasting my time on this thread when my cpu is at 2% load. 2% load while I'm sitting here typing. Why would I want my vCore to be at 1.328v right now? I wouldn't. It doesn't need 1.328v while using the Chrome browser and replying to a thread. So I changed from llc level 6 to level 7 and dropped my voltage. i can sit here all day long at 1.296v and never crash never have an issue because the very second there is any load it will go to 1.328v Ok you say your cpu is at 1.3v under load. Well lower the vCore some so where it idles at 1.26 and bumps to 1.28v under load. Because if you're doing your stress testing with the settings you have now you don't know for sure your cpu would be stable operating at 1.28v under load. But if is stable operating at 1.28v under load then lower it in bios to make it where that's what happens. If your cpu needs 1.3v to maintain stability under load then why would you want it running at 1.3v at idle? Find the lowest OPERATING vCore you can have under load. When you find what it is whether it's 1.28v or 1.3v then adjust it accordingly in Bios so that when under load that is the voltage it is getting. Vdroop means voltage will lower under load. Does it make sense to have your cpu idling at a higher voltage than it is at under load? Does it make sense for it to idle at the same voltage it needs under load? Because this is fact your cpu any cpu requires more voltage under load to be stabile that it requires while idling. Doesn;t it make more sense to have you idling voltage a little lower than your load voltage? To me it does. I don't want my cpu getting 1.328v while i'm on the internet or doing any other light task. I want it .032v under what it takes under load. Does that make sense? That is the point of overclocking. Find the lowest vCore it will take to be stable. And another thing I had a 2600k in a gigabyte motherboad and i had two options llc on or off. I updated the bios and all the sudden I had 5 or 6 levels so go to msi and see if there is a more current bios version you can update to because maybe that will be the case with that motherboard too. Anyway, hope this makes sense. I typed a long response and it got deleted so I had to type it again so I was a little more agitated this time around. But experiment with your llc settings first then ask questions because we can not suggest an answer without knowledge of your settings. We all gave you the same answer manual settings and a ver low fluctuation like that is exactly what you want IMO idle a little lower than you need under load AND YOU WILL NEED MORE UNDER LOAD

VDroop Control I would leave on Auto since your not having Vdroop. looks like the Offset is automatic on your board because the voltage goes up with load, that is a characteristic of offset, adaptive, Dynamic VID. The voltage will go up and down with CPU load variation.

Rainwater said:

Well for one man most Cpu's operate at a slightly different voltage than it is set in Bios. [...]

Click to expand...

Sorry if I've made you get upset. I didn't want to.

I'm just a begginer with OC so bear with me.

I also gave all the setting info I could get with photos.

I appreciate your help too and I'm sorry if you're message got deleted.

wingman99 said:

VDroop Control I would leave on Auto since your not having Vdroop. looks like the Offset is automatic on your board because the voltage goes up with load, that is a characteristic of offset, adaptive, Dynamic VID. The voltage will go up and down with CPU load variation.

Click to expand...

OK, then.

I'm now going to try option B with low V-droop and see how it goes.

Yesterday I did a little test playing DooM and there were no issues.

I'm aware I need to test heavier loads to ensure stability though. Maybe Intel Burn Test?