Which piece determines CPU temps?

[tincanalley]

Okay, I'm currently using a Corsair H115i Pro on my overclocked 9900K (5.1GHz, no AVX offset, 4.9Ghz cache and 1.295V).  My temps idle around 34-36c and have gone as high as 95c during stress tests.  Now I  would like to get the max temp lower, so I've been thinking of building a custom loop.  However, I have some questions as to whether this is necessary as I'm not sure my current cooling is deficient and if it is, what part of it is.  For instance, while I can't keep my max temp down below 90c during stress testing, the coolant temp never goes above 40c.  So if after an hour or so of stress testing, the coolant is only about 5 to 6c above where it started.  This makes me think the capacity of the AIO is enough, but I'm thinking it's the plate on the block that's the issue.  I don't think it is pulling heat away from the processor as fast or efficiently as possible.  So if that's the case, then a custom loop with a better water block would be called for, but not a larger radiator or higher pressure/speed fans.

 

[Blai5e]

35 minutes ago, tincanalley said:

Okay, I'm currently using a Corsair H115i Pro on my overclocked 9900K (5.1GHz, no AVX offset, 4.9Ghz cache and 1.295V).  My temps idle around 34-36c and have gone as high as 95c during stress tests.  Now I  would like to get the max temp lower, so I've been thinking of building a custom loop.  However, I have some questions as to whether this is necessary as I'm not sure my current cooling is deficient and if it is, what part of it is.  For instance, while I can't keep my max temp down below 90c during stress testing, the coolant temp never goes above 40c.  So if after an hour or so of stress testing, the coolant is only about 5 to 6c above where it started.  This makes me think the capacity of the AIO is enough, but I'm thinking it's the plate on the block that's the issue.  I don't think it is pulling heat away from the processor as fast or efficiently as possible.  So if that's the case, then a custom loop with a better water block would be called for, but not a larger radiator or higher pressure/speed fans.

 

Check to see if the cooler is loose. When I ran a 115i (I still use it on my test bench), I found that I needed to add washers underneath the 4 backplate stand-offs to allow the cooler to be snug when tightened down. Another way to achieve this is to use slightly shorter bolts as the supplied bolts bottom out prior to cranking down the cooler to the IHS. If Corsair (or more correctly Asetek) made the backplate without having the nuts closed off, this problem wouldn't be an issue (i.e. like an EKWB backplate).

[For Science!]

My bet would be TIM and IHS. 

[tincanalley]

19 minutes ago, Blai5e said:

Check to see if the cooler is loose. When I ran a 115i (I still use it on my test bench), I found that I needed to add washers underneath the 4 backplate stand-offs to allow the cooler to be snug when tightened down. Another way to achieve this is to use slightly shorter bolts as the supplied bolts bottom out prior to cranking down the cooler to the IHS. If Corsair (or more correctly Asetek) made the backplate without having the nuts closed off, this problem wouldn't be an issue (i.e. like an EKWB backplate).

It is pulled down tight.  I know what you mean about the back plate being loose.  I was worried about that when I first put it together, so I dry fitted it to see and it made contact fully and was able to snug it down.

[tincanalley]

3 minutes ago, For Science! said:

My bet would be TIM and IHS. 

This is my second 9900K.  The first one wouldn't clock above 4.9 and remain stable and there were a couple of cores that ran much hotter than others.  This second one is clocked at 5.1 and stable and all the cores are pretty equal in temps.  As a whole, this second one runs cooler than the first.  I'm guessing the TIM is better on this current one.

[AsetekDennis]

11 hours ago, tincanalley said:

It is pulled down tight.  I know what you mean about the back plate being loose.  I was worried about that when I first put it together, so I dry fitted it to see and it made contact fully and was able to snug it down.

Pretty much this.

 

Our retention kits are so, that if you tighten the thumbscrews all the way down, the correct amount of force needed for optimal performance is applied to the coldplate.

We do a lot of research on retention mechanisms and how to make sure they apply the correct force to the CPU for optimal heat transfer  

 

Now what can often be an issue is a couple of things:

 

1) TIM / Thermalpaste application is uneven or not providing full coverage

2) Thumbscrews are only "fingertight" or otherwise loose. You're more likely to break the thread of the thumbscrew than to overtighten to the point you break something (don't be stupid, of course)  

3) IHS is simply not level (several YouTube videos show how you can benefit greatly from a perfectly level IHS)

4) Contact between die and IHS can be sub-par (delid) - however, that's less relevant here as the 9900K is a sautered piece

 

Best,

 

Dennis

@AsetekDennis

[Blai5e]

10 hours ago, AsetekDennis said:

Pretty much this.

 

Our retention kits are so, that if you tighten the thumbscrews all the way down, the correct amount of force needed for optimal performance is applied to the coldplate.

We do a lot of research on retention mechanisms and how to make sure they apply the correct force to the CPU for optimal heat transfer  

 

<snip>

 

Best,

 

Dennis

@AsetekDennis

That's well and good but it I can't see that this could account for motherboard PCB's of varying thicknesses in all cases. I've had both NZXT & Corsair AIO's (a.k.a. Asetek derivatives) that exhibit loose backplates even though the cooler's bolts were screwed down till they stopped (due to the closed off captive nuts in the backplate). Both benefited from adding one or two nylon washers between the backplate stand-offs and the motherboard exhibiting a 3~5°C drop in temperature after the modification. I read of member on the EVGA forums who had to do this after using a De8uer direct die kit with an EVGA CLC before he had any real contact between the CPU and the cold plate.

 

If you've got a backplate that you can move/wobble when the bolts are threaded all the way down to the stops, I don't see how this could remotely be classified as achieving the correct force for optimal performance. I found the backplates (left stock) loose on the MSI Z270 Carbon Pro, MSI Z370 Carbon  Pro and the EVGA Z370 Classified K and as such, needed the washers for reasonable tension. For my test bench that uses a Corsair H115i, I shortened the bolts (on a belt sander & cleaned up the threads with the correct size die) so I didn't have to fiddle with washers at all.

[tincanalley]

23 hours ago, Blai5e said:

Check to see if the cooler is loose. When I ran a 115i (I still use it on my test bench), I found that I needed to add washers underneath the 4 backplate stand-offs to allow the cooler to be snug when tightened down. Another way to achieve this is to use slightly shorter bolts as the supplied bolts bottom out prior to cranking down the cooler to the IHS. If Corsair (or more correctly Asetek) made the backplate without having the nuts closed off, this problem wouldn't be an issue (i.e. like an EKWB backplate).

I really liked my 2011v3 socket for this very reason.  With the integrated bracket, there was no way to have a loose bracket.