H115i Strange Behaviour

I have recently noticed that my three year old H115i is behaving oddly.

 

If left to it's own devices, my CPU temperature is going perhaps ten degrees higher than it used to in the past.

 

When I look at this under iCEU, everything seems normal, but if run the pump for a minute or two in Extreme mode the CPU temperatures fall by that ten degrees, but when I go back to Quiet mode the temperatures stay low.

 

Incidentally I have also noticed a few times that iCEU crashes. It just sits there reporting all fan speeds and stuff incorrectly. Restarting it does no good, it requires a PC restart to get it working again.

This is somewhat similar to what we were talking about in a earlier post. As time goes by, you may get some build up in critical areas that slow the coolant flow and make the overall cooling less efficient. For anyone not using a large surface area, high core count CPU, you generally with see 0-1C difference between pump speeds (except the ultra low 1100 rpm on the Pro series). However, with enough fluid restriction the lower speed ~1900 rpm may no longer be enough to get the fluid through the block fast enough and you get some heat transfer back. Essentially the water is getting too warm while in the block because it spends too much time in there and it passes the heat back to the CPU.

 

So why would it not immediately heat up again when you set it back to the Quiet/Low speed? Possibly it takes quite a while for that 10C build up of the coolant to happen. If you leave it long enough, it may go back. Another possibility is the blockage/obstruction gets pushed out of the way on Extreme and the problem is temporarily solved by boosting the pump speed. However, eventually it comes back around and gets in the same troublesome spot causing the issue. It's hard to say from this end, but you might want to take a controlled and methodical look at it.

 

Try running a mild CPU stress test with a fixed load, like the "Bench Test" in CPU-Z. That is an easy one to run and create an even CPU wattage. Run it with iCUE up and on the cooler tab so you can see H115i Temp. When you start the program, the CPU temp should immediately jump +30-45C based on your voltage -- and then hold right there. Everything after that is down to the cooler. +1C H115i Temp should equal +1C CPU temp. You'll be able to watch the CPU and coolant temp slowly tick up 4-6C over the next 5 minutes. After 5 minutes, kick the pump speed up to Extreme. Did you get an immediate H115i and CPU temp drop of several degrees? If yes, then you are probably in the early stages of performance decline.

It sounds like this, yes.

 

Under load the temperatures ramp up when I ran a particular application and they sit there at about 55C CPU. But then an hour later its 65C still running the same application. If I switch the pump to Extreme the temperature rapidly falls to 55C again ( literally almost instantly ) . If I then turn the pump to Quiet mode again it then takes several hours to rise again to 65C.

 

I am not particularly bothered if the cooler is failing, I will just replace it in a few months time. I was thinking about doing that anyway, because I want to change my case and the new case can only takes a 360x120 radiator, so I need a new cooler anyway. In the mean time I will just keep a close eye on it.

 

Thanks for the answer!

Edited September 6, 2020 by Pocah

Yes there is definitely something very wrong. I ran the test again today. Under normal load there is not really anything noticeable. Under test load though the temperature fairly suddenly jumped from 55 to 75 degrees. Switching the pump to extreme and it fell again instantly to 55. Ten fifteen minutes later I suddenly heard the fans straining and it was back up to 75 degrees.

For now I will leave it in Extreme mode for the pump while I sort out a replacement.

Now that I have run it for an hour in extreme mode I have also noticed that the same was happening with the CPU idle but I never noticed. In fact even in idle the temperature jumps about ten degrees after a little time. The only reason I noticed this is running in extreme and allowed to full cool the idle temperature is way below what I am used to. I guess this happened over time and I just didn't notice.

Edited September 6, 2020 by Pocah

UGH! I think I may have found the issue although I am really struggling to understand it.

The head was loose on the CPU. I tightened up the screws and the problem has gone. In fact I think I had forgotten how good the cooler was, the CPU is now running the same application about thirty degrees C cooler. It's about 45 now. Why it produced those weird results, I have no idea. I don't understand why it seemed to be affected by the pump speed. However, the problem has gone. It's perhaps possible that vibration had something to do with performance, I don't know. As I said, I struggle to understand it.

Edited September 6, 2020 by Pocah

This is why it's important to monitor the coolant temperature (H115i Temp) in addition to CPU temp. Flow rate issues will always affect the coolant temperature. There are multiple reasons for the CPU temp to be higher than expected if the coolant temp isn't changing.

Nope. It seems I have not solved it. This morning I ran the application. Unfortunately I wasn't paying attention for about half an hour, but when I did the temperature was back in the 70's.

 

So further investigation.

 

I repeated the same situation. With some interesting results. ( I think my initial data was incorrect! ).

 

Running the application, the water temperature steadily increases. I get an overheat situation when the water temperature gets to about 40 degrees C.

 

Why it is increasing is a mystery though, the fans are responding, everything seems to be working as it should. The CPU is under load, but the cooler used to be able to keep everything under 50C previously.

 

Swtiching the pump to Extreme causes an immediate water temperature drop. Within a minute the temperature has dropped from 40 degrees to 30.

 

I assume that the water temperature sensor is in the head? So to me that immediately says that the problem is circulation in the head. Something is blocking the flow and stopping the water reaching the radiators, but the pump is able to overcome the problem when turned to max.

 

Does this sound reasonable?

 

My thoughts are to leave the pump on extreme until I can replace it.

Swtiching the pump to Extreme causes an immediate water temperature drop. Within a minute the temperature has dropped from 40 degrees to 30.

 

That is the clear piece of evidence you don't have a lot of time left. If you are still under warranty, you can contact Corsair for a replacement. Otherwise, you will likely want to make a decision about comes next and run the pump at Extreme.

There was a thread on the H115i and in that thread 3 years ago it said to leave the H115i pump in extreme. That is how I have mine setup and it has been that way since day one from install. I know that these AIO have a life span of about 5 years. If your pump block came loose most likely your thermal connection to the CPU from the aio pump block ..Ie the thermal paste separated too and could have dried up on the parts that were no longer touching.

 

Do you have a stethoscope for automotive use ? You can use one to listen to your pump they are pretty cheap like 9.99 at napa or even less at like harbor freight. you can listen when under a simulated load IE bench test. at normal, aka balanced load setting, quiet setting, extreme setting , you want to see if the sound of liquid is changing as it passes thru the pump and block assembly. plus if there is a restriction in the passages from trash left over from manufacturing breaking free in the radiator and making its way to the block/ pump assembly.

 

 

c-attack laid it out there too that you could be facing a failure that there is no return from. So if in doubt change it out. There are so many new better AIOs on the market now. Better to change a 100 dollar part than to have to replace a mobo and cpu in a worse case. Plus if you are restricted there is the possibility of overheat which can lead to over pressure in a sealed and closed loop which may or may not leak and that old saying goes water and electricity do not miss to well..Your system your call..

In retrospect I don't think the head came loose, rather I think I just tapped it when I inspected it the other day and that may have freed any blockage. The experiments just don't match what you would expect from "poor heat conduction". When the water temperature is low everything works just fine. It's only when the water temperature starts to rise that the problems happen. No matter what the fans do they can't stop the water temperature rising. That's not an issue with heat conduction, rather it just seems that the warm water is not reaching the radiator at all.

 

I have been running the pump on Extreme for a few days now and I still see the problem. It's not as bad as it was, but it is still there. Certainly running the pump on Extreme stops it getting out of control.

 

Anyway.... I have decided to replace it with a temporary air cooler while I return the H115i to Corsair. In all honesty I was about to throw it out anyway, I intend to replace the case soon and the case won't accept the 280 cooler, so I have to buy a new 360 instead. But I thought I might as well just buy an air cooler today because it will be interesting to do some comparative testing.

 

 

If the air cooler performs well enough I may just leave it on there and forget water cooling altogether. Thing is that it really seems to me that the pump on a water cooling system is a real weak point and when they do eventually fail, and they will, then you just have to replace the entire unit.

Edited September 14, 2020 by Pocah