Jump to content
Corsair Community

Poor Cooling Performance?


amgala

Recommended Posts

About 5 years ago, I bought an H50 cooler for my 1st gen i7 build. I ran a mild overclock and the cooler worked great. I finally upgraded my computer this year when my previous mobo died. Since I loved the H50, I thought to get another Corsair AIO for the new build.

 

My new processor is a Skylake 6700K i7, and I purchased the H100i GTX to go along with it. I installed it as per the included instructions, but for some reason it's not cooling my CPU as I would expect. Thus far I have been running at 100% stock speeds, no overclocking. I did decide to change the fans to Cougar fluid dynamic bearing units since I had a couple spares.

 

Here are my baseline readings:

http://i.imgur.com/U0J5wME.png

 

Here are my readings after running Prime95 for a few minutes:

http://i.imgur.com/Nvz1Mpj.png

 

Increasing the fan speed 300 RPM didn't seem to make much difference:

http://i.imgur.com/dbV1Oy4.png

 

My temps probably would have gone even higher, but I shut down the torture test to avoid CPU damage.

 

The rad is top mounted as an exhaust. While I would prefer to run it as a front intake, unfortunately my current case set-up does not allow me to. However, I would think even mounted as a top exhaust, the cooling should be better.

 

What do you guys think? At the highest point, there was nearly a 50 degree difference between the CPU temp and the block temp. Should my block temp be higher, meaning more thermal transfer from the CPU, and therefore lower CPU temps?

Link to comment
Share on other sites

I realized I had the pump set to Quiet mode during that test, so I changed it to Performance mode and re-ran the test. It certainly performed better than before, but after 15 minutes, CPU temps were again beyond 80 degrees C:

http://i.imgur.com/SWHyc3i.png

Link to comment
Share on other sites

Retry with a different stress test. AIDA64, OCCT, Intel XTU, anything really other than Prime. Using Prime on these newer processors with adaptive voltage (Auto is also adaptive) leads to to some scorching returns. It can be done, but requires some custom settings. Prime just isn't the universal tool it used to be. It's also a good idea to run a second type of test to differentiate between the type of test used and a potential physical problem.

 

In your test, your water temperature is staying fairly low (35-37C). This is only 6-7C over your ambient water temperature. That is a good indicator the cooler is working well. That is also the variable you are directly affecting with fan speed. The fans aid in removing heat from the water. There is nothing you can do about increasing the heat transfer from CPU die through the lid, plate, and into the water stream. That's all down to the materials used and their physical properties. I suspect Prime is causing a fairly high effective Vcore load --- higher than it should. You might be able to see this in another monitoring program.

Link to comment
Share on other sites

Retry with a different stress test. AIDA64, OCCT, Intel XTU, anything really other than Prime. Using Prime on these newer processors with adaptive voltage (Auto is also adaptive) leads to to some scorching returns. It can be done, but requires some custom settings. Prime just isn't the universal tool it used to be. It's also a good idea to run a second type of test to differentiate between the type of test used and a potential physical problem.

 

In your test, your water temperature is staying fairly low (35-37C). This is only 6-7C over your ambient water temperature. That is a good indicator the cooler is working well. That is also the variable you are directly affecting with fan speed. The fans aid in removing heat from the water. There is nothing you can do about increasing the heat transfer from CPU die through the lid, plate, and into the water stream. That's all down to the materials used and their physical properties. I suspect Prime is causing a fairly high effective Vcore load --- higher than it should. You might be able to see this in another monitoring program.

 

Wow, you weren't kidding! Thank you for the suggestion. I just ran the Intel XTU CPU stress test for 15 minutes. I then ran the RAM stress test for 5 minutes, which also maxed out the CPU load. Max CPU temp was recorded at 70 deg C, but it fluctuated a lot. The average was probably around 60 deg C. This certainly eased my fears. I didn't get a screenshot at 70 deg C, but I did manage to get one at 68 deg C:

 

http://i.imgur.com/QpVpq52.png

 

Thank you again for your help!

Link to comment
Share on other sites

You likely can reduce things even further (if necessary) by locking down a specific adaptive voltage in the BIOS, whether you intend to raise the CPU frequency or leave it as is. The Skylake processors tend to gobble up some extra voltage on the Auto setting when faced with any kind of sustained 100% load, like what is found in a stress test. Just not as much as in Prime. You aren't likely to feel the effects in your normal use, but if you are interested in tuning it further, you can tighten a few settings to keep voltage from swinging so high in these conditions.

 

Either way, the water temperature (H100 GTX temp) looks good. You can compare that to your ambient water temperature before applying a load/game/whatever to see if you need/want more fan speed. It also tells you how much you can theoretically reduce the water temps. So, in this situation even if you ran the fans as fast as possible, you could only reduce the water temperature by another 5C (at most). Most people have 5C to spare and prefer sub-maximal fan speeds. One thing to keep in mind now that you are monitoring water temperature is the your baseline starting CPU temperature is also the water temperature. So, if you raise the room temperature by 10C, the water and all CPU temperatures at the same load will go up 10C, including idle.

Link to comment
Share on other sites

Hate to necro this thread, but what thermal paste did you use and did you put a pea sized dot, a line...etc when installing it. I only saw you said installed per instructions, first thing I thought when i kept seeing the coolant temp being low even when cpu core temps were thru the roof was you had very bad transfer from the cpu heatspreader to the cpu block...maybe try only putting a pea sized dot or a line half the size of the heatspreader (and about the width of a grain of rice) and see if that helps temps much.
Link to comment
Share on other sites

Hate to necro this thread, but what thermal paste did you use and did you put a pea sized dot, a line...etc when installing it. I only saw you said installed per instructions, first thing I thought when i kept seeing the coolant temp being low even when cpu core temps were thru the roof was you had very bad transfer from the cpu heatspreader to the cpu block...maybe try only putting a pea sized dot or a line half the size of the heatspreader (and about the width of a grain of rice) and see if that helps temps much.

 

Hi veloz,

 

Sorry, I didn't see this post until now. My H100i GTX comes with pre-applied thermal paste, so I didn't have to apply any myself. Do you think I should clean off the stock thermal paste and re-apply? I do have some Noctua NT-H1 thermal paste.

Link to comment
Share on other sites

No, the pre-applied is just fine and chasing 1C around by trying to make the perfect TIM application is not everyone's cup of tea. Besides, skill in application usually trumps any minute differences in chemical material. I like NT-H1, but save it for the next time you have to take the pump off -- for whatever reason. Then it will be time to clean and re-apply.
Link to comment
Share on other sites

No, the pre-applied is just fine and chasing 1C around by trying to make the perfect TIM application is not everyone's cup of tea. Besides, skill in application usually trumps any minute differences in chemical material. I like NT-H1, but save it for the next time you have to take the pump off -- for whatever reason. Then it will be time to clean and re-apply.

 

Will do, thanks!

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...