Offset Voltage...

[Zen Ninja]

I am able to OC to 4.2 Ghz at 1.15v which keeps my temps at idle around 28-30C (H100 cooler) and 55-65C at 100% load (in the 60's only for the single hot core--there's always 1).

However, when using offset for 4.2 Ghz it throttles up to 1.28v under 100% load and only goes down to 1.025v at 1.6 Ghz (SpeedStep/C1E). I used a -0.06 and it's stable but my temps are worse--60-70C under load and 30-33C idle.

Is something wrong with my 3570k?

I'm using Prime95 BTW. Mobo is P8z77v-Pro.

Thanks in advance.

 

[Kenmitch]

You should be able to switch the offset to - values. Take a look and see. LLC will mess with you also at lower overclocks on some boards. Maybe try dropping it down if you have it cranked up. Reason being is using the - offset also lowers the idle vcore so high LLC makes you use a larger - offset.

 

[Zen Ninja]

Switched to MEDIUM (25%) LLC with OPTIMIZED Phase Control.

Offset of -0.075 for 4.1 Ghz; it's stable but runs hotter at 100% load than my own OC.

I just don't see how offset is better--I would need to set my offset to -0.150 to get my manual voltage but it would screw up my idle voltage.

 

[2is]

Why are you using LLC at all for 4.1GHz? I'm at 4.3 with the same board and chip with no LLC. Turn it off and just bump up the offset as needed for mild overclocks like that.

 

[Zen Ninja]

I pushed up to 4.3 Ghz with an offset of -0.06. Stable according to Aida64 Extreme (Full Version). I did need to set LLC to 25% though.

 

[2is]

Instead of using LLC, just bump up (instead of down) the offset. It will achieve the same results under load without compromising voltage during idle states. LLC is also known to cause momentary voltage spikes when going from full load to idle states. There's a reason why vdroop is part of intel's spec, and while it may be necessary to minimize/eliminate it for extreme overclocking, it certainly isn't for the speeds we're talking about here.

 

[Zen Ninja]

Thanks for the help, however, I think I'll keep it as is for now. I torture-tested the hell out of the OC with IBT (20 passes), Aida64 (2 hours) and now running Prime95 (coming up on hour 4).

I appreciate the advice but I don't want to bother with this anymore lol.

Max temps BTW--70C on IBT (with my one hot core at 81C max-point) and 60-63C on Prime 95 (one hot core at 72C max-point) and 50-55C on Aida64 (one hot core at 67C at max-point).

 

[Yuriman]

Well, when you do return to it, as others have said turn off LLC for such small overclocks and don't use a negative offset. LLC + negative offset = to much voltage under load and BSODs at idle.

 

[Zen Ninja]

What kind of offset then? Positive? I tried it--voltage was way too high.

 

[2is]

Not with LLC off it isn't. You can get identical results without the voltage going too low at idle/low load states which it will so with a negative offset + LLC. LLC puts far more strain on the CPU (and vrms)

 

[Zen Ninja]

Turned it off--still needed a - offset, albeit a much more modest one (-0.015). Gives the same voltage that LLC 25% and -0.060 was giving me--for both the downclock and Turbo.

Is that normal? With LLC off and a +0.005 offset my voltage for 4.3 is 1.30v (which is far too high) and idle is 1.1v (also too high).

*Note: My idle is set to 1.9 Ghz in Windows--I found 1.6 Ghz sluggish so I slowly upped the setting until I reached one that wasn't so bad (1.9).

Thanks BTW.

 

[2is]

Interesting. I'm at +.020 and my full load (IBT) voltage is 1.216 while my idle voltage is 1.016

If its stable its stable though. Definitely better off with a higher offset (even if that simply means a not so much negative in your case) than using LLC.

 

[coffeejunkee]

Turned it off--still needed a - offset, albeit a much more modest one (-0.015). Gives the same voltage that LLC 25% and -0.060 was giving me--for both the downclock and Turbo.

Is that normal? With LLC off and a +0.005 offset my voltage for 4.3 is 1.30v (which is far too high) and idle is 1.1v (also too high).

*Note: My idle is set to 1.9 Ghz in Windows--I found 1.6 Ghz sluggish so I slowly upped the setting until I reached one that wasn't so bad (1.9).

Thanks BTW.

Hmm, I use +0.015 offset for 4.2 Ghz which gives me a minimum of 1.128 under LinX load and 0.728 idle. For 4.3 Ghz I use +0.065 resulting in a minimum of 1.176 under LinX load and 0.768 idle.

All C-states enabled, no LLC. For 4.3 GHz I increased cpu power limit to 110%.

My numbers make your numbers look a bit weird so I think there's something going on that shouldn't be. I would begin with restoring the 1.6 Ghz idle speed. Increasing it in Windows is useless because EIST would already increase frequency when needed. Sluggishness shouldn't be related to the cpu imho.

Also, I have found LinX/Prime95 to be useless for stability testing, because even after passing long tests succesfully there would still be WHEA errors in Windows event viewer.

 

[Ed1]

You all probably have different VID@4.3 even with LLC off and MB probably matters too .

If you know what your stable at with Vcore@max clock under load then you can figure the offset .

Vcore - VID = offset voltage

At least it should get you close ,ballpark area .

 

[Idontcare]

Also, I have found LinX/Prime95 to be useless for stability testing, because even after passing long tests succesfully there would still be WHEA errors in Windows event viewer.

LinX/Prime95 are indeed stability testers, but do not confuse stability testing with validation testing.

Validation testing confirms each instruction in the ISA (there are >700 instructions) performs correctly, does the math right, does not generate computational errors.

Stability testing merely confirms the CPU doesn't lock up, hang, or otherwise cause such a critically fatal computing error as to reboot the system when performing calculations with a very small subset of the total ISA. (LinX probably tests maybe 30 instructions at most)

You can't have validation without being stable, but you can be stable while failing validation.

The problem with enthusiast OC'ing is that we are technically amateurs using amateur tools, whereas the engineers at Intel and AMD are professionals equipped with the expensive tools needed to fully validate a processor at any given point on the shmoo plot.

So we rely on freely available stability testers like LinX, Prime95, F@H, etc to confirm stability...but we are blind to validation.

The only way to have some level of insight into whether or not your CPU would pass validation at a given OC setting is to do things like what you are doing, taking note of errors generated by other programs, programs that access a larger (or simply different) set of instructions than the limited set used by the stability tester programs themselves.

When it comes to overclocking, as laymen enthusiasts we have absolutely no way of knowing whether our OC'ed rig would pass validation (hence the silent data corruption risk and concern, hence the advice to never OC a work computer or anything that is critical) nor do we even have a way to test and claim the rig is stable.

All we can do is use stability test tools like LinX and Prime95 to confirm when our rigs are unstable, but we can't prove they are stable (let alone valid).

 

[Zen Ninja]

Any ideas as to why my voltages are so high? Is there a setting I'm missing maybe?

Phase control?

 

[Zen Ninja]

Also, what's the problem with a negative offset if it brings me in the same range as the necessary manual voltage for stability?

 

[Kenmitch]

Also, what's the problem with a negative offset if it brings me in the same range as the necessary manual voltage for stability?

Nothing at all....As long as it's stable then you'll be ok. Use IBT to do some 1 thread, 2 thread testing also to make sure you don't BSOD under light loads just to make sure. If so a little vcore bump might be needed.

 

[Idontcare]

Also, what's the problem with a negative offset if it brings me in the same range as the necessary manual voltage for stability?

The negative offset is also applied to the idle voltage, but the idle voltage will already be reduced per the VID table...so you risk having a VID1-offset value that is stable for the OC at load but creating a VID2-offset value that is unstable for the CPU at idle.

 

[Zen Ninja]

Understood but my idle voltages are still higher than most.

For example, my idle (at 1.6 Ghz) is 1.02v--I'm fairly certain I can run the OC at 1.0v but I'll keep it at 1.02v and 1.21v (4.2 Ghz). I'm pretty sure I could lower it and I might next week but for the weekend I just want to enjoy Far Cry 3.

I will test individual threads with IBT--thanks for the advice.

 

[Kenmitch]

Just play away with it. Real world use is sometimes a better stability test than stress test apps.

IDC didn't say you'd have issues with your offset voltage, just that it's possible that you may.

 

[Zen Ninja]

Fair enough--thanks.

 

[Idontcare]

Yeah, Kenmitch is right. It is just something to be aware of if you happen to notice your computer seems to randomly reset itself or randomly hang when idle overnight and so forth. It happened to me, thats how I came to be aware of the issue.

 

[coffeejunkee]

You all probably have different VID@4.3 even with LLC off and MB probably matters too .

If you know what your stable at with Vcore@max clock under load then you can figure the offset .

Vcore - VID = offset voltage

At least it should get you close ,ballpark area .

Yes, but I think the difference is too big. TS adds +0.005 and gets 1.30, I add +0.065 and get 1.18.

And indeed, the offset is what's added to or substracted from the auto value.

When it comes to overclocking, as laymen enthusiasts we have absolutely no way of knowing whether our OC'ed rig would pass validation (hence the silent data corruption risk and concern, hence the advice to never OC a work computer or anything that is critical) nor do we even have a way to test and claim the rig is stable.

All we can do is use stability test tools like LinX and Prime95 to confirm when our rigs are unstable, but we can't prove they are stable (let alone valid).

I agree with you. But the thing is, I overclocked a bunch of Core2 cpu's and if they could succesfullly survive a couple hours of Prime95 the system would be stable for all my purposes. For my i5 750 I mainly used LinX, and again, if it could pass 20 runs the system would be stable for all my purposes. Now with my i5 3570K I get weird errors/crashes no matter how long I test with Prime or LinX. So I don't bother with them anymore beyond a very quick 10 minute test to get a rough idea of stability.

Any ideas as to why my voltages are so high? Is there a setting I'm missing maybe?

Phase control?

No, but I would start with running at stock values and checking vcore under different loads (1,2,3 & 4 threads). Next overclock to say 4GHz on auto vcore to see what the mobo supplies in that case. I haven't found the phase control option to make much difference.

 

[Idontcare]

I agree with you. But the thing is, I overclocked a bunch of Core2 cpu's and if they could succesfullly survive a couple hours of Prime95 the system would be stable for all my purposes. For my i5 750 I mainly used LinX, and again, if it could pass 20 runs the system would be stable for all my purposes. Now with my i5 3570K I get weird errors/crashes no matter how long I test with Prime or LinX. So I don't bother with them anymore beyond a very quick 10 minute test to get a rough idea of stability.

It is unfortunate that my programming capabilities are so extremely limited, and I'm not just trying to be humble there as I really live on a bubble when it comes to my coding capabilities, but it is clear that we enthusiasts could really use a software tool that (1) heated the processor up as LinX does, and simultaneously (2) executed computations on every single instruction in the ISA in various permutations (like memtest86+ does) while checking results to validate operation across the entire ISA at that temperature, voltage, and clockspeed.

Even better would be if this program would be able to tap into the BIOS to drive an auto-tune overclock program that mapped out the shmoo plot for any given unique combo of a consumer's CPU/mobo/HSF/ambients so we just launched the program and walked away.

Come back and you have a pretty graph that shows you your stable clockspeeds versus voltage, temperature, and CPU power usage to choose from. And then the BIOS would absorb those value and implement a truly customized turbo-clock profile that ramped the entire spectrum of multipliers.

If I could program I would endeavor to create such a utility. Alas, I suck in this dept