H100i V2 not cooling very well

Hi all,

 

recently completed a new build with a 7700K CPU on an MSI Z270 Krait Gaming mobo with an H100i V2 cooling the proc. my previous build had the original i7 920 with an H80 cooler, which worked really well even with a fairly serious oc in place.

 

On the new rig the cooling doesn't seem to be as efficient so I've checked setup (CPU fan set to manual 100% power, Corsair link is running and shows pump and fan speeds. CPU block is screwed in correctly and seems tight).

 

So have tested today giving about half an hour for the system to reach a stable idle temperature then used the prime 95 small FTTs (max heat) and large FTTs (max power) for over 30 minutes each and found that on max power test the package temp stabilised at around 70C with the coolant at around 30C, on the max heat test it stabilised at around 82C with coolant at around 31C.

 

Just wondering if this is 'normal' from other experience or if this seems like a bad contact and a reseat, maybe with some washers on the back (which has been suggested in a couple of places)?

 

Would be grateful for any thoughts on this.

Actually, I would say the opposite. You are doing better than most other Kaby Lake owners with those tests, which are always going to be hotter on the socket side. This is one of the changes on newer CPUs and I could run Prime for days on my old 930. I won't even use it on my HW-E. However, to really know if the heat is transferring into the unit, you need to look at the coolant delta, from pre-test stable level to the peak. What was the starting H100i v2 Temp at the start of those tests? If you didn't take note, just use your normal baseline idle now. That should be relatively constant, save the daily fluctuations in room temperature.

 

I would hold off on the washers until there is a proven issue. You don't want them if you don't need them. You also might consider running a stress test that is less physically contained by the CPU. Prime and IBT are really too intense to use a cooling system check. You will always have more heat than the CPU and lid can transfer out through the cold plate and into the coolant stream. You can use Linpack tests, OCCT, AIDA, even an easy test like Intel XTU for a longer duration will tell you more about your cooling system than Prime.

Hi,

 

Thanks for the quick reply.

 

Meant to add that detail. Pre-test temps were around 35C-40C for the package and about 28C-29C for coolant. This was why I was concerned because the coolant only goes up by about 2C-3C.

I would run a milder test of any sort, one you can stomach for longer. When you run Prime, the CPU is going to hit 80C+ in 1-2 seconds. Most people are going to chicken out (wisely) or the CPU will throttle before long. Heat accumulation in the coolant stream is usually a product of time. (You add more than the fans can remove). The problem with Prime on newer CPUS is it adds more heat to the CPU than the CPU can transfer it out. The fan speeds and coolant temps don't play much of a factor. It's all on the CPU side.

How long are you thinking? The max power test that was around 70C was running for 55 minutes previously and that temp was pretty stable for around 50 minutes of that time. All I'm glad of now is that my combo boiler is more efficient at heat transfer or the start of my day would be freezing :biggrin:

No, that’s plenty. The longest I ever run synthetic tests for is about 30 min with me nearby. I am completely against the old mentality of a “24 hour stability test”, unless that is the way your pc will be used - like a server or mining rig. You can use your normal programs to make final adjustments. A long gaming session may produce higher case and coolant temps anyway. Even 10 min is enough to assess cooler functionality. The longer you run, the more case design, environment, and other related factors influence the result. I don’t think you have an issue, but a 10 min run of something else should make you feel better.

Your temps are right in line with what I see from my 7700K at stock. A big part of the problem is that around the Haswell era, Intel switch to using a pretty crappy TIM between the core and the heat spreader. There's nothing that a cooler can do about that.

Now, if you are brave, you could de-lid your processor and replace the stock TIM with a better TIM. I saw a good 20C drop by doing this and using Cool Laboratory Liquid Ultra. It will, however, most definitely void your CPU warranty and there is the potential of killing your CPU.