4770K, the nightmare continues?

[Stick]

Some useful CPU heat managment ideas here ...

 

ISSUE WITH MX-4

 

I OC'd mine to 4.4 GHz in Nov 2013, and over many months of heavy duty number crunching, I saw overall temps go higher.  So 6 months ago, I delidded, removed the dried out IntelTIM, and applied MX-4.  Overall, all temps went down about 10C but recently I am back at +100C on core#0 again with an app that originally ran at 80-85C.

 

So what happened with the MX-4?  On a 2nd careful delid, I found that silicone had separated from the gray particle matter, so half the die was covered in pure oil while the other half in semi-dry gray material.  That's very bad news for MX-4 in my opinion. 

 

For decades I've used white Thermalcote #250 electronics TIM.  Depending on application, I'd change the consistency by desorbing its oil between two sheets of printer paper.  The Thermalcote never dries, even after months of paper oil desorption.  It only gets thicker and stays pasty.

 

But for CPUs I found Thermalcote not to be as efficient as MX-4, when fresh, now I know.   So I tried the same desorption trick on MX-4 (half-pea blob squashed to a silver dollar diameter) and to my great surprise, it completely dried out in just 5 minutes, leaving flaky dry gray powder! 

 

Clearly this experiment agrees with the MX-4 delid oil separation between die and IHS.

 

So I mixed the dry MX-4 power with fresh Thermacote 50:50 and got a consistency slightly thicker than original MX-4, but now with the fabulous slicone base of Thermalcote and thermal transfer characteristic of MX-4. I applied it to the CPU and fan, and I'm back to the same cool temps as originally in November.

 

UNBALANCED TEMPS

 

I have had the same problem of unbalanced core temps across a 4770K when all are running 100% with IntelBurnTest which I at first thought were related to TIM.

 

The data below was taken after a new application TIM mixture (to delid and fan) yesterday, as described above.

 

My application.  Typically I see a 6C differecence between core#0 and core#3 when running around +80C overall. 

Core #0 83

Core #1 81

Core #2 76

Core #3 77

 

IntelBurnTest.  That difference increases to about 13-14C when I approach 100C overall.  

Core #0 100

Core #1   97

Core #2   89

Core #3   86

 

Note that core#1 also runs much higher.   Hence core#0 vs core#3 difference increases with inceased CPU load. 

 

ProKon said:  "I have had similar results with the temps increasing on the delid over time, however judging by the large difference in core individual temps from your chip I would have to say you used to much \ too little mx-4 or did not apply the tim very evenly."

 

From these results I think the problem is separation and drying out of MX-4, not its application. Also, the core0-core3 temp difference to me does not appear related to TIM, no matter what you use (even Intel TIM stock I saw the difference).  Thus there must be some other gremlin responsible for the temp difference, such as cores getting (even slightly) different mobo voltages or a die step/fab manufacturing issue, since it seems ProKon did not report a difference problem like you and I are seeing.

 

The bottom line is .... all these TIMs are organic.  At the high temps we run, these materials will degrade, dry out, or separate out, as experienced here only after a few months of operation.  Nothing I, you, or Intel even with its new 4790K, apply can beat the old metallic bond of die-to-IHS, a lost art in favor of greed.

 

How will the MX-4 powder + Thermalcote solution last?  I have some high hopes, but, we'll see with time.

 

 

[BiG StroOnZ]

If you are going to Delid, use a tried and proven TIM that has always been used as well as recommended to be used for Delidding; like CoolLabs Liquid Ultra or Liquid Pro.

 

I don't see why you would keep using the same TIM repeatedly after it exhibits the same problem time and time again. It's obviously an issue with MX-4. 

[patrickjp93]

A real shame intel decided to fudge the damn IHS up, Haswell could have been equal to Sandy in OC headroom.

No it couldn't have. The process shrink combined with some of its architectural qualities (meant to save power) automatically hinder it. Broadwell is going to be worse still, even if soldered.

 

On a separate note, have you tried putting a level on it and seeing if the surface really is level? If it isn't, you can use sand paper to lap the surface of the heatspreader until it's level. Use the figure-8 method for best amateur results (I use a series of Sharpton stones normally used for sharpening culinary knives and straight razors, but that's probably way outside your budget).