H80i v2 the water is cool but the CPU is hot

I have had this system for probably 4-5 years and I haven't been watching the temperatures, but the other day I was. At idle the CPU is running around 22C and the H80i around 27C (I assume that is the temperature of the water). Reported using Corsair Link 4. See my profile for info. Running stock speed. Yes, the water is hotter... at idle!

When I push the CPU with Prime95 (max heat) the CPU goes up to 90C but the water temperature (well what Corsair Link outputs) only goes up to about 34C. For tests like this I have the fan on max (about 1500rpm) and the pump on performance (about 3000rpm) just to be the best cooling.

I was confused how the water could be so COOL but the CPU so HOT! Since the water temperature changes and the fan speed and pump speed changes (using Corsair Link) I assume those are working. Well, I can clearly see the fan spinning. One tube is warmer than the other so I assume there is some water circulation. Perhaps not ideally working, but nothing obvious...

My only conclusion was that the water pump cooler wasn't mounted well on the CPU. Got some Artic MX-4 and removed and replaced it. Temps pretty much the same. When replacing the cooler I noticed some sloshing of water and was concerned, but I'm assuming (?) this is standard. The pump is quiet so I don't believe it's pumping some air or cavitating.

While perhaps not a top of the line water cooler, I'd expect it to do better than this and when I built the system I assume I had much better temps because I do remember testing it!

Any insights would be greatly appreciated!

Thanks

Nothing you described here is out of the range of expectation.

Prime95 is brutal on a CPU and its use of AVX instructions does generate a lot of heat. So you'll see those temps, especially on a non-delidded Skylake. Stock settings on many motherboards are also a little overly happy about pushing the voltage up, which increases heat more.

Liquid has a high specific heat (or heat capacity) than the CPU and the silicon. It takes significantly more heat energy to raise the temperature of the liquid 1C than it does to raise the temperature of the CPU 1C. Additionally, heat transfer isn't 100% efficient, especially (as mentioned before) if your CPU hasn't been delidded.

Now, the H80 doesn't have a lot of liquid in it (so lower heat capacity of the system) and it has a small radiator (so less surface area to release waste heat) so it'll get warmer under extended load than a larger radiator would. Additionally, if you have the radiator mounted as exhaust (and that's the typical installation for an H80), then you will also be pushing waste heat from other components (such as the GPU, VRM, etc) in the system through the radiator, which can actually warm the liquid and certainly decreases overall cooling efficiency.

Thank you for your information. I'm still concerned that a water cooler like this can't get the temperatures down lower. Perhaps as you said, this CPU just "runs hot" and the H80 isn't a super AIO cooler.

Yes, I have it mounted with the typical exhaust configuration and I run the the stock BIOS settings. However since I don't see the water getting that hot, I assume radiator is managing to cool it down adequately. I'm just still puzzled why the water doesn't get hotter when the CPU is so hot!

Perhaps a higher end air cooler would be better? I'd like to see the temps in the 70C range or so with Prime95. I don't have room for something like the H100.

+1C coolant temp = +1C CPU temp and the same in reverse. I am guessing you don't have a +15C coolant rise under load, so the cooler will never be able to reduce the CPU temp by 15C. It can only reduce the temperature baseline back to the starting point basis.

If you take a fixed load like Prime 95 or perhaps a milder version like the CPU-Z Bench Stress test, start the test with no intro or warm up period. Note the coolant temp and then the jump after the load is applied to the CPU. The difference between the baseline coolant temp and the that flat line load CPU temp is your coolant to CPU differential. This gives you an idea how much you can let the coolant go up before you hit your limit. If you are already at your limit at the start, you will have to turn down the voltage. No cooler can reduce that differential. It is all conductive and an assessment of the CPU's own heat level.

Even with custom watercooling, multiple rads, dozens of fans, and powerful pumps, you'd have similar results. That's how CPUs behave.

They generate a lot of heat on a very small area, so they will always read hotter than other components with lower density (GPUs).

An example from my own rig : at full load, my CPU may run at 70° @150W, while the GPU would never go beyond 45° @250W.

Radiators also are more efficient at removing heat when the difference of temperature is higher between ambient air and coolant. So, the coolant idle temperature will always stay somewhat "hotter", because it's extremely hard to cool say 27° water with 23° air.

In general, air coolers have better idle temps, but watercoolers have better load temps.

All that to say, your H80 works as it should :) Prime95 is okay to test stability, but totally not realistic for max temps under usual loads. That's OCCT levels of Barbecuery. Maybe try a looped cinebench to get a better idea?