Why is my i5-4670k so hot at only 4ghz?

KaaN10

Honorable
Oct 26, 2012
15
0
10,510
First build and overclock, I've just been changing the cpu ratio and voltage via the Z87 bios. I also changed the RAM to 1600 before starting if that makes a difference.

The i5-4670k is 4ghz @ 1.125 voltages (lowest it will go before BSOD). The temps I'm getting are 75 degrees max on prime95 after a 30 minute blend test. The cooler is a hyper 212 evo with a noctua f12 fan.

Any ideas why it's running so hot for only 4ghz and 1.125v?
 
for a start, you shouldnt need to up voltage at only 4ghz. I am running an IB 3570k at 4ghz on a hyper tx3, and it probably gets to 70c all 4 cores loaded up, ambient temp around 20c. They are quite hot chips due to the crappy thermal paste between the cpu and the heat spreader. You can de-lid the cpu and replace that thermal paste, but its a bit risky
 
Actually, from reading on the subject, it would seem the thermal paste is not the issue with the Ivy Bridge and Haswell processors, but rather the fact the IHS has been glued, and the glue causes there to be a larger gap that the paste needs to bridge, which paste is not very suitable for regardless of it's quality.

I agree about adding voltage unnecessarily. Was it done because the chip was unstable at 4 GHz or because it was only thought to be needed?
 

Zac Lloyd-Jones

Honorable
Jul 4, 2013
247
0
10,760
My 4670K is running at 4Ghz at 1.1v and got to a max of 47 degrees in an Aida64 stress test. I am using the same Cooler Master Hyper 212 Evo but I am using the bog standard fan that came with it. Yours really seems a bit hot. :S
 

Intel God

Honorable
Jun 25, 2013
1,333
0
11,460


The TIM between the die and IHS is the issue with Haswell and Ivy. Sandy used Fluxless solder which is why it runs so damn cool. Before i delidded at 4.7Ghz i would hit 96-97C in cinebench. After deliddding we're at a cool 59C at 4.8Ghz with the same test.
 

Zac Lloyd-Jones

Honorable
Jul 4, 2013
247
0
10,760


Wow Intel God that's one hell of a difference. What cooler are you using?
 


did you just de-lid, and leave the die exposed direct to heatsink? or did you just relace the TIM with something better and re-seat the lid? This is something i would consider doing, when there is need for more horsepower, which i have no need for yet......
 

Intel God

Honorable
Jun 25, 2013
1,333
0
11,460


I've tested it both ways. Delid and remove all the sealant around the IHS and replaced the Intel TIM with Arctic MX4 gave me an 18C drop. Running now with just the waterblock directly cooling the die gave me a 38C drop. With my 3770K i used "Coollaboratory Liquid PRO" between the die and IHS and it dropped me 29C but you cant do that with Haswell because it's conductive so i decided to just run IHSless and cool the die directly.

My CPU cooler is an EK block to the left of my post.
 
I appreciate the heads up about Sandy Bridge, and already am aware the IHS is soldered on.

There is an extra amount of space between the IHS and silicon on both Ivy Bridge and Haswell which is caused by the glue used to secure the IHS to the processor package. That gap is too great for the TIM to effectively pass heat between. Did you replace the glue yourself after you removed the IHS, or just tossed in new TIM? If you did not recreate the gap, you did not do a correct comparison between your TIM and what Intel originally used.

For all we can speculate, whether it's bad TIM, bad application of the TIM at the factory, or an excessive gap, it may also just be a clever method by Intel of artificially limiting the range in which average enthusiasts run and overclock their parts, as the risk involved in removing the IHS is not something everybody is going to be willing to subject their processor to.
 

Intel God

Honorable
Jun 25, 2013
1,333
0
11,460
If you look at the bottom of the IHS after delidding you can see there is ample contact between the die and IHS. IMO its the TIM

fe5mny.jpg
 
I would like to disagree with you as respectfully as possible. Below is a link to an article where an Ivy Bridge CPU is measured before and after having it's IHS removed and TIM replaced. The measurements are taken with and without shimming the IHS to the same height as stock, to determine whether or not it's the TIM or excessive gap the TIM is made to span. The conclusion is actually that Intel's original TIM performed better than the aftermarket TIM. However, when the lid is allowed to rest directly on the die, with no shim, the aftermarket TIM performs as one would expect. :)

The conclusion here is that too much of a gap exists between the silicon and IHS for the TIM to work effectively.

The article can be found here:

http://forums.anandtech.com/showpost.php?p=34053183&postcount=570
 

Jake Anderson

Honorable
May 31, 2013
156
0
10,690


Its a combination of both the TIM and the glue.

The glue does cause a slight gap between the IHS and the DIE, thus making heat transfer less efficient, when CPU's were soldered you had a complete metal contact from DIE to IHS although the gap was still there.

TIM is less conductive than solder, thus again you lose the efficiency that you had with the solder. So the slight gap and the TIM both make it worse. As you stated when you ran MX-4 you only had 18c drop but with Coollaboratory Liquid PRO or you can use Ultra, you had 29c, thats because of no gap (same as with MX-4) and now a liquid metal TIM..

BTW, you can use Coollaboratory Liquid PRO or Ultra or any other liquid metal TIM with Haswell.. I have CLP on my 4670k.. You just have to cover your VRM with a non conductive/ capacitive TIM, or clear coat polish first.

http://www.overclock.net/t/1313179/official-delidded-club