4790k abnormal temperatures with nh-d15

A

aaabcbaa

New Member
I bought this 4790k for my new rig about a month ago, and immediately overclocked it to 4.6 ghz @ 1.23v. The system seemed stable during prime95 stress test;however, all 4 cores shot up to 100c after 15 minutes (when the fft length was reduced to 8k). I attempted to resolve the problem by checking and later reseating the heatsink, but the temperature was still hot enough to boil water.:eek:

Is there any possible fix to address this issue? Or did I receive a defective chip?
Feel free to ask me for any info if needed:)

(psu: evga supernova 1000p2, case fans: crappy fans that came with the corsair c70 case)
 
Don't be using the stock heatsink/fan that came with the cpu. You will need an aftermarket cpu cooler when overclocking. The coolermaster hyper 212 EVO would work nicely.
 
He is using a Noctua NH-D15 (they're beasts).

I get a similar behavior with my 4770K and D14 but it isn't delidded. What kind of temps do you get after 30-60 sec?
 
johnb35 said:
Crap, missed that in the title.

Thermal pad or paste not correct?
Click to expand...

I've tried both arctic silver 5 and nth1, and reseated the heatsink 3 times:mad::mad::mad:, but as you can see the temperature still is scary at only 1.23v. Do I need to consider delidding my cpu and apply some liquid metal between the die and the IHS?
 
how much paste are we talking? plus if your using prime run it on blend not FFT. Prime isnt the best software to use especially on fft's. I personally use Aida64 it still stresses the cpu above normal limits but not the the point where it nearly sets on fire like prime does.
 
What are your ambient temps?
Are the case fans Corsair or some unknown brand cheap ones?
 
How much paste are you applying? Is the cooler fitted with fans and all working. If it has 2 are they in push pull?
 
tylerjrb said:
How much paste are you applying? Is the cooler fitted with fans and all working. If it has 2 are they in push pull?
Click to expand...

I used the dot method for nt-h1 and the vertical line method for the arctic silver 5. I cannot use the recommended push-push configuration for the nh-d15 due to compatibility problems with my ram, but I later reversed the fans to a pull-pull configuration (not sure how this would affect the performance), and the airflow remained the same.

Another problem that I have noticed is that using Prime95 would result in far more BSODs and overheating. 4.5ghz (1.18v) with occt was at least 10 degrees cooler than with prime95.

I've redone a test at 4.7ghz, 1.27v (for 5 minutes)
And here is the temperature graph: (use the link)https://drive.google.com/file/d/0B4-X8UpTtPy7UFhuNjA0NmZUaUE/view?usp=sharing
As you can see, the temperature spikes shoot up to 97c, but it generally stays below 85c, is it healthy for the cpu?

This is a different graph resulted from a test at 4.7 ghz, 1.26v (This frequency is Prime95 unstable, but "everything else" stable) https://drive.google.com/file/d/0B4-X8UpTtPy7QXd2ejJ0QnNEbFk/view?usp=sharing
But as you can see, temperature at this voltage is a lot more consistent, and I cannot understand how this is happening.

aaabcbaa said:
This is a different graph resulted from a test at 4.7 ghz, 1.26v (This frequency is Prime95 unstable, but "everything else" stable) https://drive.google.com/file/d/0B4-X8UpTtPy7QXd2ejJ0QnNEbFk/view?usp=sharing
But as you can see, temperature at this voltage is a lot more consistent, and I cannot understand how this is happening.
Click to expand...

Third 4.7ghz stress test result, 1.26v, time span 10 minutes (to prevent inconsistency in results). Highest spike 95c, but generally stays well below 85c. Graph: https://drive.google.com/file/d/0B4-X8UpTtPy7N2hpc2hIRjV4em8/view?usp=sharing
 
Last edited by a moderator:
Technically, it shouldn't go over 70 degrees C. Either its a bad processor, bad temp readings or heatsink isn't applied correctly.
 
Last edited:
Do you have them in pull pull like two fans ontop of each other or a fan on either side like
Airflow-->heatsink<--airflow as this will cause problems. Have you tried using the stock Intel cooler and seeing what temps you get?
 
You shouldn't use Prime95 on Intel CPUs newer than Sandy Bridge, as I've read it can pull more voltage than you've allowed.
I run 2600K @ 1.275v with D14 and never see temps above 55-60C.
Yours should be a little warmer if run at the same voltage, but your D15 should be a little colder :)
 
Newest update: problem solved, my chip was defective.
I listed my cpu on craigslist and bought another one from ncix, the temps never went over 85c again in occt (4.6ghz, 1.23v) :)
https://drive.google.com/file/d/0B4-X8UpTtPy7dmFjeXlEWnY2am8/view?usp=sharing I stressed it for 30 minutes, and over 90% of the time the temperature was below 80c.

My assumption: since intel has discontinued soldering the IHS to the die on non-enthusiast grade cpus, it was possible to buy a cpu with below-average thermal conductivity. Delidding is probably a mandatory procedure to prevent Haswells from frying themselves.
 
Last edited:
Even with that cpu it should be well under 80c my old 3770k did 4.8ghz at less than 80c at 1.4v, I even ran it at 1.55v 5.1ghz once and it hit 90c on prime. and the cooler performance and cpu temperature wise is similar. If you have two fans pushing towards each other like you stated previous it will reduce performance as they are pretty much holding the heat in. I'd make sure to atleast have either one fan only or both in push pull. Your temps will drop.
 
tylerjrb said:
Even with that cpu it should be well under 80c
Click to expand...

It's pretty well known that Haswell used a pretty crappy TIM between the core and heatspreader. I can hit 90C pretty easily on Intel Burn Test at mild clocks. While the Devil's Canyon refresh changed the compound it's still not particularly great.
 
But it shouldn't be over 70 degrees c correct? If it was me, I wouldn't push it over 65c.
 
You must log in or register to reply here.
Back
Top