3570K Not Overclocking Well? 1.25Vcore @ 4.4Ghz

[geokilla]

Seems I need 1.25Vcore to get to 4.4Ghz under load. That's a 300Mhz bump in speeds, at a cost of 500mV of Vcore. (my CPU overclock in sig may be outdated). I guess my chip has degraded? Temps have maxed out at 80C during specific tests in Prime95, but otherwise it seems to normally top out at around 75C. I'm currently stress testing it for stability.

Is 1.25Vcore the absolute max I should go? I'm don't mind running the CPU at 75C to 80C under Folding@Home load. No program ever stresses a CPU out as much as Prime and Intel Burn Test does anyways.

 

[john3850]

With ibt 100% load I need 1.3v for x47 and 1.264v for x44
If I use a fixed multiplier no x16 idle my vcore will be less.

 

[tweakboy]

Great voltage on that OC. gl

 

[geokilla]

Seems I need 1.25Vcore to get to 4.4Ghz under load. That's a 300Mhz bump in speeds, at a cost of 500mV of Vcore. (my CPU overclock in sig may be outdated). I guess my chip has degraded? Temps have maxed out at 80C during specific tests in Prime95, but otherwise it seems to normally top out at around 75C. I'm currently stress testing it for stability.

Is 1.25Vcore the absolute max I should go? I'm don't mind running the CPU at 75C to 80C under Folding@Home load. No program ever stresses a CPU out as much as Prime and Intel Burn Test does anyways.

 

[24601]

With Intel Math Kernel Linpack I can get 4.6ghz @ 1.29v, but it throws WHEA errors every week or so so i backed down to 4.5ghz @ 1.25v.

It will also depend on what you are using to cool it, your case (and it's cooling) and the ambient temperature though.

 

[monkeydelmagico]

I'm running almost identical figures to yours. 1.246v 4.4ghz. Using prime and large fft setting runs my cpu at 78 degrees C. I have a small air cooler though, xigmatek 92mm and I'm running the fan at a set 1000 rpm's for noise reduction, so take the temp results with a grain of salt.

 

[john3850]

If I ran Folding@Home 24/7 I would go with a fixed multiplier and keep below volts below 1.3v.
I also got WHEA errors at x46 and x47 once the volts are are over 1.31v no more WHEA errors.
It seems you have two choices a 44 multiplier with a lower vcore which many of the smarter people use or a 46-47 multiplier and a +1.3v.
Because I use wc I choose 46-47 multiplier and a +1.3v. on my 3770k.

 

[tweakboy]

You cant get 4.4Ghz with a 1.25v....... For 4.4Ghz you need 1.35v or 1.38v

 

[geokilla]

If I ran Folding@Home 24/7 I would go with a fixed multiplier and keep below volts below 1.3v.
I also got WHEA errors at x46 and x47 once the volts are are over 1.31v no more WHEA errors.
It seems you have two choices a 44 multiplier with a lower vcore which many of the smarter people use or a 46-47 multiplier and a +1.3v.
Because I use wc I choose 46-47 multiplier and a +1.3v. on my 3770k.

How are your temperatures?

 

[john3850]

As a rule I never watch the temps.
Right now with a game in the back round and ibt running standard or 1024mb only 1 core hits 83c with 1.288v at 4500 right now with a 3x120 radiator with 6 100cfm fans.
The cheap Apogee water block has a bad habit of clogging do to the tiny pins in it.
My i5 runs the same stuff at a higher load then the I-7 do to no ht.

 

[tweakboy]

John3850 you have soo many fans soon that pc will get wings and fly away.

Must sound like a Jet in your room.......unless their low rpm. gl

 

[walshlg]

May I suggest looking into de-lidding. There are a lot of folks that have found that the thermal compound connecting the die to the ihs (can) is very very badly applied. If you are game, there are vids on utub on de-lidding, you then just clean and re-apply thermal compound to the die and stick the can back on and proceed as usual. Certainly consider this if your cores show significantly different temperatures from one another.

 

[bunnyfubbles]

overheating VRMs are what held back my 3570K more than anything else (starting to wonder just how many people are getting throttled by overheating VRMs without realizing it - as I've experienced this with my X79 rig as well - because you have to spot it during testing, as the system can easily be stable and pass the overnight stress tests), although CPU temps were also starting to get out of control and I would have needed to delid if I was going to push it beyond 4.6 (ended up settling for 4.5 due to VRMs)

 

[FalseChristian]

I've been running an 'old' i5 2500K at 4.5GHz using 1.325v for years without 1 problem. Any thing under 1.4v is fine with a good after-market CPU cooler.

 

[MrSquished]

been running my i5 3570k at 4.4ghz for i guess like a year now. totally stable.

under full load with prime95 Coretemp shows my Voltage maxing out at 1.1609v

when i reboot ill check in the bios what the voltage is set at. watching monday night football on espn.com so i don't want to reboot now

 

[BonzaiDuck]

May I suggest looking into de-lidding. There are a lot of folks that have found that the thermal compound connecting the die to the ihs (can) is very very badly applied. If you are game, there are vids on utub on de-lidding, you then just clean and re-apply thermal compound to the die and stick the can back on and proceed as usual. Certainly consider this if your cores show significantly different temperatures from one another.

If I'd built an Ivy Bridge system, I would've felt compelled to do that. Some people had really good results over a short-term of reporting with Liquid Ultra or similar indium-gallium-based thermal compounds. Others who used regular replacement TIM material seemed to show short-term regression to the thermal equivalent of the INTEL paste. There were worries that the Liquid Ultra or Indigo Xtreme TIM might cause damage to the processor as Intel might have anticipated with their soldering process, but I never heard anything (anyone else?)

I've been running an 'old' i5 2500K at 4.5GHz using 1.325v for years without 1 problem. Any thing under 1.4v is fine with a good after-market CPU cooler.

Well, the issue is the lithography -- the closeness or "packing" of circuits -- and electromigration from over-volting. I think my old Q6600 system was volted to near 1.38V -- fixed for an overclock. At that time, Intel published a "safe range" and a wider "operable range," and the upper bound of the safe range was about 1.375 -- if I recollect properly. The lithography (I think . . ) was something like 45nm. Then Sandy Bridge came along, with a 32nm lithography; collective wisdom said 1.35V. It should only be lower -- maybe 1.30 or 1.32V -- for Ivy @ 22nm -- and don't quote me!

But things have also changed since OC'ing Yorkies, Kentsfields and earlier processors. You can over-clock with precision while leaving EIST, C1E etc enabled. So -- unless you're "Folding" or running marathon benchmarks, the processor is not going to sustain the higher voltages for very long, and the residual risk is then simply in the load-to-idle transition spikes -- so brief that the monitoring software won't even register them.