6600k overclock temps - need suggestions

[samfisher770]

Hi everyone. I assembled by 6600k rig yesterday and overclocked it to 4.2GHz through the motherboard's bios based overclocking utility.

Here's the config -

i5 660k with CM Hyper 212 evo cooler (1 fan)

gigabyte z170x gaming 7

g skill 2 x 8 GB ddr4 2400MHz

Windows 7 64 bit

After a few minutes of Prime 95 (small FFT test), the temps are going up to 89C. I have attached few screenshots for reference. Could someone advise if I got a really bad piece or otherwise.

The room temp is around 35C. As for real life usage temps, it went up to 63C after playing CoD Advanced Warfare for about 20-25 mins.

Any suggestion would be appreciated.

[Energycore]

10 minutes ago, samfisher770 said:

Hi everyone. I assembled by 6600k rig yesterday and overclocked it to 4.2GHz through the motherboard's bios based overclocking utility.

Here's the config -

i5 660k with CM Hyper 212 evo cooler (1 fan)

gigabyte z170x gaming 7

g skill 2 x 8 GB ddr4 2400MHz

Windows 7 64 bit

After a few minutes of Prime 95 (small FFT test), the temps are going up to 89C. I have attached few screenshots for reference. Could someone advise if I got a really bad piece or otherwise.

The room temp is around 35C. As for real life usage temps, it went up to 63C after playing CoD Advanced Warfare for about 20-25 mins.

Any suggestion would be appreciated.

 

The 212 EVO isn't a good CPU cooler. You're probably going to have to take the voltage down, if you just used the voltage your motherboard offered you can probably go down.

 

Another solution might be to get some high performance fans (2 of them) to put in there.

[Bananasplit_00]

try putting the voltage a bit lower, should help with temps, if you just used the motherboards auto OC then the voltage is porbably higher then it needs to be

[samfisher770]

Thanks for the inputs. As for voltages, any suggestions regarding what voltage range I should target initially? Or should should I start with 0.05V decrements?

[Lawliet93]

Yeah, lower the voltage, 1,27 is to much for 4,2GHz, you should be able to achieve it at 1,15V, but, don´t take my word fr it, do some stability tests and make sure, that you are running the lowest voltage possible. And... you could try  deliding the CPU. It has been shown, that ivy bridge, haswell  and skylake CPUs have crappy thermal paste under the heatspreader. Use something like arctic MX4, or thermal grizzly hydronaut. 

[samfisher770]

Delidding the cpu would be more than i can handle. as for stability tests, are prime95 tests the only benchmark for stability? If no, please suggest some others.

[Imakuni]

11 hours ago, samfisher770 said:

Thanks for the inputs. As for voltages, any suggestions regarding what voltage range I should target initially? Or should should I start with 0.05V decrements?

Mine needs 1.296v to achieve 4.3ghz. Yours seems very much in line with what I'd expect for 4.2.

 

At any rate, if you want to try and lower it, go changing it by as little as you can at a time. Also, don't forget to set LLC properly - my voltage within Bios is actually set to 1.27, but it goes up to 1.296 after LLC quicks in.

 

Oh, and btw, keep cache at stock speeds, at least while you figure your core OC.

8 hours ago, Lawliet93 said:

It has been shown, that ivy bridge, haswell  and skylake CPUs have crappy thermal paste under the heatspreader. Use something like arctic MX4, or thermal grizzly hydronaut.

While some of these use worse TIM than the others, this isn't so much the TIM's fault as it is the heat spreader and the laws of physics...

[samfisher770]

update - the coretemp readings are probably not of core voltage, as cpu-z reading shows much much lower voltage reading.

I have also measured the reading using a multimeter and they almost match with the readings from cpu-z.

So the question now, is whether these Vcore and Vid are okay. Please suggest.

 

[Imakuni]

52 minutes ago, samfisher770 said:

update - the coretemp readings are probably not of core voltage, as cpu-z reading shows much much lower voltage reading.

I have also measured the reading using a multimeter and they almost match with the readings from cpu-z.

So the question now, is whether these Vcore and Vid are okay. Please suggest.

Normal. This can happen with Skylake, as the voltage regulator is on the mobo, not on the CPU. Use HWMonitor:

[Tylerr]

which prime are you using?

 

try version 26.6

[Imakuni]

10 minutes ago, Tylerr said:

which prime are you using?

 

try version 26.6

Why? That version is potato and doesn't properly use the CPU.

[samfisher770]

During the tests, Vcore goes much lower (1.14) than what I set it in bios(1.22). Should I increase the Vcore a little, if the Vid of 1.32 (approx, as reported by coretemp ) is okay?

[Imakuni]

6 minutes ago, samfisher770 said:

During the tests, Vcore goes much lower (1.14) than what I set it in bios(1.22). Should I increase the Vcore a little, if the Vid of 1.32 (approx, as reported by coretemp ) is okay?

I suppose your LLC isn't set properly. My voltage too drops below specified if I let LLC on Auto; fiddle with it a little to adjust the voltage.

[samfisher770]

Finally got to 4,2 GHz stable @ 1.165V. During Prime 95 tests, it goes down to 1.152V (DMM reads 1.142 though) even with LLC set to high.

Ran Prime 95 v28.9 for 5.5 hrs straight - no issues whatsoever. This is what I will settle for now. Thanks everyone for all your inputs.

 

[Tylerr]

On 6/6/2016 at 4:24 PM, Imakuni said:

Why? That version is potato and doesn't properly use the CPU.

 

yeah it does

 

http://www.tomshardware.com/answers/id-2790414/high-temperatures-skylake.html#16586288

[Imakuni]

30 minutes ago, Tylerr said:

yeah it does

 

http://www.tomshardware.com/answers/id-2790414/high-temperatures-skylake.html#16586288

As I said, version 26.6 doesn't properly use the CPU.

 

You see, Haswell~Skylake have an instruction set called FMA3, which is only used by more modern versions of P95. The guy you've quoted forgot to mention that this is actually the reason why it it gets so hot: the software is squeezing every possible ounce of performance out of the chip, by making use of all the latest features it has.

 

Sure, if you want to go potato and not use your processor to the best of it's capabilities, by all means use an early version. There's nothing wrong with that.

 

But if you want to use the CPU to it's fullest, then you have to use newer versions of Prime95, as those can awake the hidden processing power inside that little piece of silicon. There's absolutely no problem in doing so, and that's what people should set as baseline in my opinion (which OP seems to share).

[Tylerr]

5 hours ago, Imakuni said:

As I said, version 26.6 doesn't properly use the CPU.

 

You see, Haswell~Skylake have an instruction set called FMA3, which is only used by more modern versions of P95. The guy you've quoted forgot to mention that this is actually the reason why it it gets so hot: the software is squeezing every possible ounce of performance out of the chip, by making use of all the latest features it has.

 

Sure, if you want to go potato and not use your processor to the best of it's capabilities, by all means use an early version. There's nothing wrong with that.

 

But if you want to use the CPU to it's fullest, then you have to use newer versions of Prime95, as those can awake the hidden processing power inside that little piece of silicon. There's absolutely no problem in doing so, and that's what people should set as baseline in my opinion (which OP seems to share).

the only thing they do is give you unnecessarily high temperatures.

 

if you pass the test with 26.6 version you'd pass it with the newer ones, they'd just run hotter.

[TheGamingBarrel]

On 06/06/2016 at 9:24 PM, Imakuni said:

Why? That version is potato and doesn't properly use the CPU.

It is used because it does not have AVX2, Which puts UNNECESSARY Stress on the CPU, barely anything uses it.