Higher temps after 4790K delid

[BlueWeasel]

My current setup is a 4790K running at 4.7ghz (1.27v) cooled by a Thermalright True Spirit and 2 x 140mm fans (push-pull).

With this setup, I was getting temps greater than 80C during Prime95 testing. I delidded the 4790K, reapplied Noctua NT-H1 between the die and IHS, as well as between the IHS and heatsink. Initial tests looked good after the delid, with max temps dropping 70-72C...almost a 10C drop.

Now a week later, I ran the same test and the temps on cores #2 and #3 shot up to 85-88C! I'm seeing a huge variance in temps:

Core #1 70C
Core #2 87C
Core #3 88C
Core #4 82C​

I did everything by the book and didn't apply much thermal paste. Prior to delidding, the variance in temps was much less and they were all within a few degrees of each other.

Any idea as to the problem? I'm guessing a gap between the die and IHS. Maybe I didn't put enough TIM, as I put much less than the original Intel material.

 

[Durp]

Did you scrape off the black adhesive from the IHS and the CPU PCB? Some people forget this step and it's important.

 

[BlueWeasel]

Did you scrape off the black adhesive from the IHS and the CPU PCB? Some people forget this step and it's important.

Yes, I did.

 

[buklau]

I think the best way is to get liquid ultra on the IHS/die. I did the same thing to my 4790k, added MX2 and the temperature didn't drop much.

 

[BlueWeasel]

I think the best way is to get liquid ultra on the IHS/die. I did the same thing to my 4790k, added MX2 and the temperature didn't drop much.

Use Liquid Ultra/Pro between the die and IHS and regular thermal paste (the Noctua NT-H1) between the IHS and heatsink?

 

[twjr]

Use Liquid Ultra/Pro between the die and IHS and regular thermal paste (the Noctua NT-H1) between the IHS and heatsink?

That's exactly what I did with my 4690k and 3570k and had significant (15-20c) drops in temps.

 

[Yuriman]

It's possible that the IHS is not sitting centered on the die; the socket retention bracket tends to slide it when you lock into place, and this could be partly responsible. It could also be that you're experiencing something IDC was calling "pump-out". Either way, the first thing I'd do is remount with NT-H1 and see if it happens again.

 

[BlueWeasel]

I kinda get the feeling that the IHS isn't exactly flat and I didn't use enough TIM between it and the die. To be honest, I applied using the "pea method" and only put slightly more than what is recommended for typical (non-delidded) applications.

I'm considering getting some CL Ultra or Pro to try. However, it's a really thin material so wouldn't there still be gaps between the IHS and die?

 

[realibrad]

I kinda get the feeling that the IHS isn't exactly flat and I didn't use enough TIM between it and the die. To be honest, I applied using the "pea method" and only put slightly more than what is recommended for typical (non-delidded) applications.

I'm considering getting some CL Ultra or Pro to try. However, it's a really thin material so wouldn't there still be gaps between the IHS and die?

Using more compound will only help make sure that there are no gaps. Thermal compound has less thermal conductivity than the lid. What is more likely is that your lid or compound has moved due to things like vibrations. If you had not used enough TIM, then the temps would have started high. The fact that it was lower, then went up means something has moved.

Adding more TIM will fill up all of the cavity under the lid and movement will not matter there. I would suggest taking off the lid again and seeing where the TIM has settled. I bet it moved.

 

[buklau]

Use Liquid Ultra/Pro between the die and IHS and regular thermal paste (the Noctua NT-H1) between the IHS and heatsink?

That's correct. I use CLU under IHS/die, and MX2 IHS/heatsink, -20 to -25c across all cores with prime95 load. I tried using MX2 before, but will quickly degrade after 2 months and I have to re-apply TIM again to avoid temperature increase.

 

[BlueWeasel]

Now that I think about it, I think I let the IHS slide when I clamped it down.

We leave for vacation tomorrow but will remount next weekend. I'll let you know how it goes. Thanks for the advice.

 

[Sheep221]

Locking CPU in socket moved the IHS and spread paste off die area leaving some cores hotter.