[Karan98]

So I just built my rig with a new 4790K and I am experiencing weird temps. In AIDA64 everything is nice and cool, but using IBT AVX or Prime95 sends the temps up into the high 80's and low 90's?

Any help?

Thanks

 

[OS-Wiz]

Did you use the TIM (paste) pre-installed on the cold plate? Its good stuff. I found that my older H100 preferred a ceramic-based TIM.

 

[Karan98]

On the Cooler Master Nepton 280L there is no TIM preapplied, only TIM included that you apply yourself. I have remounted 2 times. It's weird because in every stress test except Prime95 and IBT things are nice and cool.

 

[Karan98]

Bump

 

[xgunnas]

I thought you aren't supposed to p95 on a Haswell.

 

[electro2u]

Set your voltage lower. 1.31v for 4.4Ghz is definitely unnecessarily high. Try 1.2 and see if it is stable. if not, go to 1.225 etc

Apply the TIM with the grain of rice method (slightly longer than an actual grain of rice, but you want it in a vertical line on the center of the chip. Then when you set the heatplate of your Nepton cooler onto the chip, give it a little shimmy/twist or 2 and then fasten the cooler as normal.

IBT and Prime95 are notorious for extremely high temps and yours are pretty normal. You can improve upon them, but it's not surprising. There are a lot of these same thread every day so lots of people experience the same.

Aida64 is a very lightweight and not particularly useful stress test. You can pass AIDA64 for a long time and then get BSOD on normal loads.
Quote:

Originally Posted by xgunnas

I thought you aren't supposed to p95 on a Haswell.

p95 is fine on Haswell but the most recent version 28.5 with AVX functions is really really extreme. It's better to use 27.9

 

[Karan98]

What about IBT?

 

[electro2u]

Quote:

Originally Posted by Karan98

What about IBT?

It pisses Haswell off ^_^
It's not very useful either but it gets very hot.
Use XTU benchmark, it is the preferred way to score overclocks.
For long term stress testing OCCT is preferred. Read the OP of http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics/0_30

 

[Karan98]

Quote:

Originally Posted by electro2u

Set your voltage lower. 1.31v for 4.4Ghz is definitely unnecessarily high. Try 1.2 and see if it is stable. if not, go to 1.225 etc

Apply the TIM with the grain of rice method (slightly longer than an actual grain of rice, but you want it in a vertical line on the center of the chip. Then when you set the heatplate of your Nepton cooler onto the chip, give it a little shimmy/twist or 2 and then fasten the cooler as normal.

IBT and Prime95 are notorious for extremely high temps and yours are pretty normal. You can improve upon them, but it's not surprising. There are a lot of these same thread every day so lots of people experience the same.

Aida64 is a very lightweight and not particularly useful stress test. You can pass AIDA64 for a long time and then get BSOD on normal loads.
p95 is fine on Haswell but the most recent version 28.5 with AVX functions is really really extreme. It's better to use 27.9

I went from auto CPU Voltage to Manual and I'm stable at 1.15V 4.4GHz. Peak temp of 77c on IBT AVX but generally hovered and averaged at around 60-63c. Using OCCT and Aida64 temps were around 57-61c. Will order some better MX4 thermal compound and see if there are any improvements.

 

[electro2u]

Quote:

Originally Posted by Karan98

I went from auto CPU Voltage to Manual and I'm stable at 1.15V 4.4GHz. Peak temp of 77c on IBT AVX but generally hovered and averaged at around 60-63c. Using OCCT and Aida64 temps were around 57-61c. Will order some better MX4 thermal compound and see if there are any improvements.

I sadi OCCT but I meant x264, sorry. It's a little less user-friendly you have to set it up properly but the OP of that thread I linked above explains it. Sounds like you have solved the issue anyway, good work!

 

[Karan98]

Quote:

Originally Posted by electro2u

I sadi OCCT but I meant x264, sorry. It's a little less user-friendly you have to set it up properly but the OP of that thread I linked above explains it. Sounds like you have solved the issue anyway, good work!

Going to run 100 loops of this overnight, check the results and see what I can get in terms of OC's

 

[Karan98]

Ran x264 overnight, 4.6GHz 1.264V under load max CPU temp was 75c cores were sitting in the upper 60's lower 70's. Good so far?