[SOLVED] Weird temperatures after replacing cooler

Jose Luis Lopes

Honorable
Apr 5, 2015
75
1
10,665
I just changed my CPU cooler from a 212 EVO to a Cooler Master Masterliquid 240, system boot up fine, but now I'm getting weird fan speeds and CPU temperatures (it's an i7 4790k btw)

My mobo, an Asus z97-K doesn't have a CPU PUMP header so I plugged the pump into the CPU Fan header and left it at 100%, and the fans to a chassis fan header with a standard curve, but I'm getting 100C reading and 100% fan speed at the minimum load of the cpu. windows login? 100%, opening chrome? 100% etc etc

The weird thing is I don't think it really is at 100C, the CPU isn't throtling remaining even at 4Ghz at "100C", even if I leave it during some time it does not blue screen or freeze, and the moment I stop loading the CPU the temperature goes from 100C to 30C IN ONE LITERAL SECOND

To check if I did something wrong with the cooler instalation I removed it to re-seat it and re-apply thermal paste, but everything stayed the same, and one thing I noticed, the CPU wasn't even hot to the touch even though it was at 100C just 1 minute before I remove the cooler. I have updated the BIOS in the meanwhile but didn't help

Any ideas?
 
Solution
Standard Corsair directions for those older boards state Pump on Cpu_fan and Fans on any sys_fan/cha_fan of your choice. It's done this way for only one simple reason, the likelihood of the pump failing is guessed at being greater than fan failure. And cpu_fan header has a factory default alarm where if it sees nothing plugged in or 0 rpm, it'll shut the pc down or refuse to boot, done to protect the cpu from catastrophic thermal overload.

Which makes sense from that position, and covers Corsair's butt if the pump dies fail.

But since the aio is controlling the temp of the cpu, having a fan which responds to case/motherboard/system temps is asinine.

Swap the connectors. Put the pump on cpu_aux or any sys_fan/cha_fan header and in...

Karadjgne

Titan
Ambassador
Standard Corsair directions for those older boards state Pump on Cpu_fan and Fans on any sys_fan/cha_fan of your choice. It's done this way for only one simple reason, the likelihood of the pump failing is guessed at being greater than fan failure. And cpu_fan header has a factory default alarm where if it sees nothing plugged in or 0 rpm, it'll shut the pc down or refuse to boot, done to protect the cpu from catastrophic thermal overload.

Which makes sense from that position, and covers Corsair's butt if the pump dies fail.

But since the aio is controlling the temp of the cpu, having a fan which responds to case/motherboard/system temps is asinine.

Swap the connectors. Put the pump on cpu_aux or any sys_fan/cha_fan header and in bios make sure it's locked to 100%. Put the fan splitter on cpu_fan. When the cpu is cool, aio fans barely move. When the cpu heats up, aio fans spin faster to compensate. That'll fix the temp reporting oddities.

As to the actual temps, that's most common from 2 reasons.

First being there's an air bubble/pocket trapped in the pump. Easy fix, unplug everything then give the whole pc a tight little shake for a few seconds to break up the air and allow it to pass on to the radiator.

Second is its not seated fully. This is common from beginner builders who get scared they'll break the cpu if it's screwed down too tight or whom screw only one corner down fully at a time. Lay the pc on its side so the pump is face up. Loosen all 4 screws slightly in an X pattern, 3 full turns should be good. Check the standoffs under the mount, they should be fully seated and very finger tight. Put slight pressure on the pump, forcing it onto the cpu and twist side to side very slightly. This'll seat the pump fully and help even out the paste. Holding the pump down tight with 2 fingers, tighten the thumb screws in an X 1 turn each, continuing until all 4 are snug, then in the same X pattern, give all 4 screws a final tweak to make sure they really are tight, or bottom out, whichever comes first.

Now boot into bios and give it 10 minutes, watch the bios temp. It should be 40-50ish °C. If that looks good, finish booting to windows and see what the temps are then.
 
Solution

Jose Luis Lopes

Honorable
Apr 5, 2015
75
1
10,665
@Karadjgne

So, changed the connectors, Pump is now on chassi fan with full speed, splitter with 2 radiator fans on CPU fans on a standard curve,
Algo gave it some shake, from what I read having the radiator vertical helps dissipate and prevent air bubbles, that's exactly how I have it, so it shouldn't be it
30 minutes on BIOS; stable at 39C, didn't move, fans at 700-900RPM

restart into windows, temp 50C on boot, but fans stable and didn't increase, put some load on the CPU, this happened
zqv65WN.png

with jet engine fans as a plus...

from 50 to 100 in 3 seconds. from 100 to 50 in 2 seconds.

I don't believe the temperature reading is correct, if it was, why wouldn't the CPU thermal throtle or even shutdown completely? I didn't get a blue screen, lag, anything, during this test he was most of the time boosted to 4.2Ghz. Is it possible that my thermal sensor is the problem? May I have somehow damaged it during the cooler installation? (and case swap btw)
 

Karadjgne

Titan
Ambassador
It's.... possible. Anything can happen, Murphy's Law. But to the cpu, that's beyond astronomically unlikely as those sensors are stragetically placed inside the cores, so have tons of protection surrounding them. And that single temp is either a single core temp, or an average of the temps, not sure how cpu-z reads it.

You could run a Prime95 large fft test, using all the workers. Simultaneously run Coretemp or Realtemp and watch each individual core temp. If all are running high and @ 5°C ±, then the sensors are fine. If 1 goes nuts, shutdown the Prime immediately. Check the paste. It's unlikely, but possible that there's a gap/bubble in the paste allowing that 1 core to overheat. Other than that, the only option I can see is a defective sensor. And your cpu is out of replacement warranty.

I've asked for some help on this, hopefully he has time, so hang tight 😁
 
Last edited:

Jose Luis Lopes

Honorable
Apr 5, 2015
75
1
10,665
@Karadjgne

So, I think it was bubbles after all. After a few more shakes I started hearing some "bubbling" (can't really describe the sound), so I left the PC running with a slight inclination over the night, and now it won't pass 80C during a stress test, which I still think it's a little too hot since it's below 15C ambient, but way better than before, hopefully it will improve a bit more over time

I did re-seat the cooler and re-apply thermal paste before posting, so I'm sure that wasn't the problem, and as you told me to check, the core temperatures never were more than 5C apart from each other, so it shouldn't be the sensors, but I'm still baffled how the computer wouldn't shutdown or throttle when hitting 100C...

Oh well, it seems fixed, thanks
 

Karadjgne

Titan
Ambassador
80°C during a stress test is pretty normal, depending on the test. I recommend Prime95 small fft, with AVX technologies disabled. Half an hour run to climatize the coolant. That's a straight 100% load that's almost exactly the same as a 100% gaming load as it uses the same instruction sets, so is an absolute worst case gaming temp. Normal gaming will be well under that, should be around 55°C ish, give or take.

Other stress testers do use some different methods, such as using ram, linpack etc and that can kind of skew the results, often creating a 80% to 130% virtual load, (which you'd never hit without doing some seriously wierd number crunching professional stuff. So basically, never), and with temps bouncing all over kinda leaves you wondering why you bothered in the first place.
 

Jose Luis Lopes

Honorable
Apr 5, 2015
75
1
10,665
80°C during a stress test is pretty normal, depending on the test. I recommend Prime95 small fft, with AVX technologies disabled. Half an hour run to climatize the coolant. That's a straight 100% load that's almost exactly the same as a 100% gaming load as it uses the same instruction sets, so is an absolute worst case gaming temp. Normal gaming will be well under that, should be around 55°C ish, give or take.

Other stress testers do use some different methods, such as using ram, linpack etc and that can kind of skew the results, often creating a 80% to 130% virtual load, (which you'd never hit without doing some seriously wierd number crunching professional stuff. So basically, never), and with temps bouncing all over kinda leaves you wondering why you bothered in the first place.

I bothered because in the beggining it wasn't bouncing, it was 40C, then 100, then back to 40, there wasn't 60 or 80 stable temperature ever, but I guess it was the air bubbles trapped in the loop. Yes the 80C was during prime95, during gaming it is now 55-70 depending on the game, so I guess it's fine now

thx for the help