High Core Voltage with I7-4770K

wouterkcs

Member
Feb 13, 2009
40
0
61
When OC to 4.3 Ghz using a GA-Z87X-UD3H with F7 BIOS using Vcore at auto with a stress test Voltage increases to 1.465. No OC Vcore is at 1.133. Similar configured Ivy bridge Voltage is 1.2. What causes this over voltage? Thanks
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Probably your motherboard. Change voltage from "Auto" to "offset".
 

Hulk

Diamond Member
Oct 9, 1999
4,191
1,975
136
Your Haswell chip might not be as good an overclocker as your Ivy too. Silicon lottery.
What are your temps at that voltage?
 

Ketchup

Elite Member
Sep 1, 2002
14,545
236
106
I thought Ivy's didn't overclock as well. Plus, as mentioned, the auto overvolt of the Gigabyte board. It is possible the Ivy could run on less if set to manual, unless you have already looked into this.
 

wouterkcs

Member
Feb 13, 2009
40
0
61
With 4.3 65C. Default at 3.5 57C.

Voltage looks reasonable up to 4.1 @ 1.261.

Anything above it spikes to 1.46 to 1.55.

Talked to Gigabyte support today and stated temperatures above about 50 are not tolerable. I thanked him and said "Good Bye". I read many are running temperatures to 70C and 80. I am using a Antec KUHLER H2O 920 -- not a bad cooler!

This board is V1.1 (new) with Bios F7 (very recently). Wondering if new BIOS has issues, or motherboard?? I assume the CPU is okay.
 

Hulk

Diamond Member
Oct 9, 1999
4,191
1,975
136
I have read that voltages over 1.5 can kill on air so be careful.
 

Morbus

Senior member
Apr 10, 2009
998
0
0
High voltages and high temperatures kill CPUs over time. Running that high a voltage you need to ensure bellow 50ºC, otherwise a haswell CPU can handle about 60ºC, more than that it's pushing.

50ºC is what the CPUs are designed for, it doesn't mean they won't withstand higher temps.
 

Geforce man

Golden Member
Oct 12, 2004
1,731
5
81
High voltages and high temperatures kill CPUs over time. Running that high a voltage you need to ensure bellow 50ºC, otherwise a haswell CPU can handle about 60ºC, more than that it's pushing.

50ºC is what the CPUs are designed for, it doesn't mean they won't withstand higher temps.

This is completely 100% false.

Intel has in their spec sheets these CPU's designed to work for years @ 99 degrees c, on the stock cooler. If you have it under that, you are good. That said, the voltage shouldn't be above probably 1.35 unless you have extreme cooling on Haswell.
 

BSim500

Golden Member
Jun 5, 2013
1,480
216
106
With 4.3 65C. Default at 3.5 57C.

Voltage looks reasonable up to 4.1 @ 1.261.

Anything above it spikes to 1.46 to 1.55.

Talked to Gigabyte support today and stated temperatures above about 50 are not tolerable. I thanked him and said "Good Bye". I read many are running temperatures to 70C and 80. I am using a Antec KUHLER H2O 920 -- not a bad cooler!
You can run at 50-70c easily long-term without issues. For long-term personally I'd aim for a maximum of 70c. 1.46v sounds too high a jump from 1.26v as if your board is still on automatic and over-estimating voltage (unless that's the "Haswell thing" of the IVR adding 0.1v when running AVX benchmarks?). Does the voltage spike up in games or just benchmarks?

Try disabling offset / adaptive and set a fixed voltage of say 1.27-1.3v and see how high it can go then. If the worst comes to the worst, 4.1Ghz is roughly on par with a 4.3-4.4GHz Ivy Bridge, which isn't "slow" on any game. Unless you're an extreme overclocker, a rather large +0.2-0.3v doesn't sound worth it to me just for an extra 200MHz.
 

wouterkcs

Member
Feb 13, 2009
40
0
61
I have another Haswell that OC to 4.5.
As I have issue with this one GA-X87X-UD3H with F7 BIOS, is the high voltage caused by?
1) CPU
2) Motherboard
3) BIOS

Thanks
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
I have another Haswell that OC to 4.5.
As I have issue with this one GA-X87X-UD3H with F7 BIOS, is the high voltage caused by?
1) CPU
2) Motherboard
3) BIOS

Thanks

Not too familiar with this latest gen of Gigabyte boards, but we have some earlier "UD" releases here.

Really, I have to say -- aware that many here have more thorough knowledge -- "auto" VCORE overclocks are not the way to go with this hardware. I use ASUS boards, and the "AI-overclocking" or parallel features always push the processor voltage much higher than needed.

Use "offset" mode. Find a feature in the BIOS that addresses "extra voltage in Turbo," although I can't be sure the Gigabyte boards offer it -- but they should . . . really . . In the case the board has such a setting, you can use a minimal or modest offset -- even positive -- and then adjust the second item.

As others here have said, the maximum temperatures for these processors are in the spec sheets. For Ivy Bridge, I think it was either 100C or 105C. The real damage could occur from the voltage -- over time -- with electromigration eating away at the processor. I think with the Ivy Bridge, we'd concluded that 1.32V was a "safe" limit. You might think for Haswell it is lower. And some may say I'm too cautious -- WHICH -- I am.
 

wouterkcs

Member
Feb 13, 2009
40
0
61
With an offset of .05 with VC set to normal and OC of 4.3 using TURBO method in BIOS resulted in stable system:
1) VCORE 1.247
2) Realtemp of 70C

Obviously it was the GA-X87X-UD3H that has an issue with run away voltage. Hope GA will fix this.

Thanks to the guys having me try the offset. THANK YOU!

Now I need to delit to get temperatures down!

Any other parameters I need to change to improve stability?
 

2is

Diamond Member
Apr 8, 2012
4,281
131
106
This is completely 100% false.

Intel has in their spec sheets these CPU's designed to work for years @ 99 degrees c, on the stock cooler. If you have it under that, you are good. That said, the voltage shouldn't be above probably 1.35 unless you have extreme cooling on Haswell.

I haven't seen that spec sheet, but even if it does exist, all of Intels specs are based on DEFAULT speeds and voltages. Once you start OCing, and especially over voting, those numbers are completely meaningless.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
I haven't seen that spec sheet, but even if it does exist, all of Intels specs are based on DEFAULT speeds and voltages. Once you start OCing, and especially over voting, those numbers are completely meaningless.

The sensors are going to work properly no matter what the speed or voltage -- unless the voltage is pushed to a point where slow deterioration of the silicon gives way to more rapid damage. But the thermal properties of the silicon are the same.

The processor either deteriorates from temperature -- stopped by the throttling limit -- or from electromigration due to over-volting beyond the recommended "safe" range within the known "operable" range.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
With an offset of .05 with VC set to normal and OC of 4.3 using TURBO method in BIOS resulted in stable system:
1) VCORE 1.247
2) Realtemp of 70C

Obviously it was the GA-X87X-UD3H that has an issue with run away voltage. Hope GA will fix this.

Thanks to the guys having me try the offset. THANK YOU!

Now I need to delit to get temperatures down!

Any other parameters I need to change to improve stability?

Interesting that your using the KUHLER H20. I was going to suggest you replace the stock cooler before I reviewed posts and saw that.

I've been away from here for a while. IDontCare's thread on delidding the Ivy Bridge opened up possibilities as well as uncertainties. I assume someone has tried delidding on the Haswell by now?
 

2is

Diamond Member
Apr 8, 2012
4,281
131
106
The sensors are going to work properly no matter what the speed or voltage -- unless the voltage is pushed to a point where slow deterioration of the silicon gives way to more rapid damage. But the thermal properties of the silicon are the same.

The processor either deteriorates from temperature -- stopped by the throttling limit -- or from electromigration due to over-volting beyond the recommended "safe" range within the known "operable" range.

Sorry, but that's simply wrong. Those specs are based off default values. There's always deterioration at any voltage, even default. Overvoltage and overclocking negates any specs intel puts out there. Wishful thinking and guess work is just that, and what you're saying is based on nothing more than those two factors.
 

coffeejunkee

Golden Member
Jul 31, 2010
1,153
0
0
With an offset of .05 with VC set to normal and OC of 4.3 using TURBO method in BIOS resulted in stable system:
1) VCORE 1.247
2) Realtemp of 70C

Obviously it was the GA-X87X-UD3H that has an issue with run away voltage. Hope GA will fix this.

Thanks to the guys having me try the offset. THANK YOU!

Now I need to delit to get temperatures down!

Any other parameters I need to change to improve stability?

It's not an issue with the board, it's just normal. Every board from every brand overvolts on auto, and the higher you oc the worse it gets. It's why you don't use auto vcore when overclocking, unless maybe for a very mild oc and then it's still not optimal.

Sorry, but that's simply wrong. Those specs are based off default values. There's always deterioration at any voltage, even default. Overvoltage and overclocking negates any specs intel puts out there. Wishful thinking and guess work is just that, and what you're saying is based on nothing more than those two factors.

Yes, but it's less wrong than what Morbus was saying. 50 or 60 degrees is nothing for a cpu and temps like that will surely not cause any meaningful degradation, unless with 'over time' he means 20 years.

Intel website is just confusing for starters as well, listing the 72 degree tcase value when no software can actually read that value. The only thing we do know is that at normal clocks and voltage 99 degrees is the thermal throttling point.

Personally I don't mind seeing 90 degrees during Linx stresstesting, normal temps will be way lower anyway.
 
Last edited:

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
It's not an issue with the board, it's just normal. Every board from every brand overvolts on auto, and the higher you oc the worse it gets. It's why you don't use auto vcore when overclocking, unless maybe for a very mild oc and then it's still not optimal.

[answering 2is]

Yes, but it's less wrong than what Morbus was saying. 50 or 60 degrees is nothing for a cpu and temps like that will surely not cause any meaningful degradation, unless with 'over time' he means 20 years.

Intel website is just confusing for starters as well, listing the 72 degree tcase value when no software can actually read that value. The only thing we do know is that at normal clocks and voltage 99 degrees is the thermal throttling point.

Personally I don't mind seeing 90 degrees during Linx stresstesting, normal temps will be way lower anyway.

I settled on my 4.6Ghz clock for this now-dated Sandy Bridge core when I saw the four core temperatures hovering around 76C under a punishing LinX test. Generally, my ambient for those tests was between 78 and 82F.

I'm only guessing that the IHS (integrated heat spreader) for Ivy and Haswell cores is not too different than it was for SB. It's copper with nickel-plate. TCase was supposedly defined to measure temperature at center, and it was always so many degrees -- maybe 10C below an average for the core sensors.

Of course, all that changes with Intel switching from a solder applied between the IHS and the silicon, to a thermal paste. With the understandable risk, replacing it with liquid ultra would again reduce the difference from a hypothetical TCase. But I'll agree with everyone else: there is some modest interaction between voltage required and temperature, but the ill effects are two different things.

Certainly, I might expect greater risk with higher temperature with the more-compact Haswell (what is it -- 16nm or something like that?)

Anyway, a 4.3 clock (if achievable) on a Haswell might mean better performance than a 4.6 or 4.7 on an old Sandy. And somewhere I and some others already remarked that we're less gung-ho on OC'ing the newer cores.

Then, there's the ability to enable the power-saving features with an overclock setting. Since additional speed also generates additional heat, but you're only running full-out at those speeds for some fraction of the time the processor is powered throughout its lifetime, the voltage damage risk would be more attributable to the spikes in transition -- unless the turbo-voltage setting is way beyond a range around default.

It seemed that the spikes were cited as a greater risk in the Yorkfield article a few years ago. With a fixed VCORE, you are just applying a higher voltage continually, and you still experience the spikes.

But again -- point of it -- these last three generations of cores were designed with a "turbo mode." They made the chip smaller and reduced the power requirement. The default clock speeds have changed very little over the three lines of processors. So it now seems less useful to OC -- just for a few hundred Mhz.

EDIT: Memory fades! That is -- my own! The lithography for Haswell is same as Ivy Bridge or 22nm? I see now [about six months tardy]. The TDP is lower at 77W -- same lithography. I'd only guess that higher voltage is the bigger risk with that. . . . If you're going to OC, you might want to impose a more severe limitation on over-volting VCORE. And [EDIT] -- maybe SOME processor models have lower TDP, but I now see the 4670K at 84.
 
Last edited: