Weird processor temperature problem

Metric Tensor

Hey lads,

Built my new-build back in September. I have an i7-6700k and a Cooler Master Nepton 240M cooler - running on an ASUS Z170 Delux board.

The computer has been running at least five days a week and sometimes six or seven since I built it in mid-September.

Sometime in October I started it one morning and as the morning progressed I could hear the processor fans getting louder and more under pressure - there was an accompanying. Initially I thought there was something stuck in the fans but I'm pretty sure the "clicking" was actually the pump going at full tilt too. So I downloaded RealTemp and lo and behold the processor was at 80ish degrees at idle! Before going to the hassle of removing the cooler and putting on new paste, etc I tried a restart and it went back down to idling at 20ish degrees and pushing up to the 50s when work kicked in.

It hasn't put a foot wrong since - until this morning - when I started up today I heard what I think was the pump under a bit of pressure again and the processor fans running faster than normal so I started up realtemp immediately and the idle temperatures were in the 40s. I was pretty busy with work so I didn't have time for too much diagnostics so I just rebooted am back to my low 20s idle temperatures again.

So those two times in the 100 odd start ups since the computer was built are the only times I've seen this behaviour. Any ideas lads?

My thoughts are:
1. Some sort of miscommunication between the MB and pump that means it's not one when it thinks it is. (Although I think I can hear it)
2. Faulty pump (Likewise I think I can hear it)
3. Some sort of mounting problem.
4. Something else that I can't think of!

frozenfrozen

sounds like a faulty pump to me

ED E

#1 Guess would be the AIO Pump is fed off the wrong header and isnt managing the spin up(the hardest part).

Connect the pump to a 12v Molex to fan adapter and see how you get on.

Metric Tensor

Thanks lads.

The header is specifically called the "W_Pump" header and the ASUS software can tell me the RPM of the pump and all. The speed can be adjusted based on processor temperature.

I can try connecting it at full speed directly the PSU or even just setting it at full speed using the MB software but the problem only seems to happen once a month so I don't know how diagnostic that will be.

Johnboy1951

You should also eliminate the possibility of some background process putting a very heavy load on the CPU - a process that is only run very seldom.

Cuddlesworth

frozenfrozen wrote: »

sounds like a faulty pump to me

A working pump doesn't click. Faint whine yes, no clicking though.

TerrorFirmer

Not sure what the click could be but the high temps may be due to airlock. Try taking the rad off and moving it around a good few times and remount it.

Metric Tensor

Thanks lads.

Definitely an audible clicking. But come to think of it I'm not sure I've heard it at any time other than on these two occasions. It was the type of noise that made me think maybe a cable tie was sticking into one of the fans. It wasn't constant though - it came and went.

I ran some stress tests on the processor previously and it maxed out in the high seventies and the pump was audible but as a little whine, not a click.

I'll stress test it again later to see and I'll take off the rad and give it a bit of a shake sometime tomorrow or at the weekend.

I'm also considering running the pump at 100% for a while to see. With the loading it gets at work the processor never gets to a temp where it would demand 100% pump speed - maybe that's a problem too.

The big problem is that it could be February before I see the problem again. Makes it very hard to do any diagnosis.

ED E

If you have another PC/Laptop to work on for a bit:

1. Add a reboot script to the startup folder
reboot.bat

shutdown -r -t 45

2. Add a temp monitor program to the start folder
3. Reboot
4. Watch the machine reboot every 45 seconds for a while, see if the temp ever spikes way high.

Cuddlesworth

Thinking about it, my AIO and ASUS mobo have a weird bug where the pump runs 100% out of sleep. And once it just simply turned off.

Metric Tensor

Thanks for all the comments lads.

I ran Prime95 on a torture test this evening just to see what would happen. I watched the temperature of each core using realtemp and the fan and waterpump speeds using the Asus AI Suite.


The core temps went up to mid to high 60s but the fan speeds didn't change and the "CPU Temperature" reading in the ASUS software never went above 40 degrees!

So I took matters into my own hands and pushed all the fans and the pump up to full speed manually. This drove the peak temperature down a small bit to low 60s.
Just when I changed all the controllers to full speed there was a short noise that sounded like a little crackle. A bit like the noise you would get from an old HDD when it was chugging away thinking about things. That lasted less than a second and all I could hear was the fan hum then, but as I pulled the profiles back down to "standard" again I heard the same little noise. (This crackling is what I earlier referred to as clicking but I heard it over a slightly longer period.)

I've just tested it a few times and it replicates every time I switch the whole set up between max fans and standard. I am not sure if it comes from the pump or not ......I'll do some more checking and update tomorrow.

Thanks again.

Metric Tensor

Waaahaaaay - a bit of good news. The crackle is coming from the one of the fans - not the pump.

When I increase and decrease the pump speed there is no noticeable change in noise. But when I increase and decrease the CPU fan I get the noise.

I'm not as concerned about the noise now. The crackle could be a fan bearing or something.

Of course I still have the random heat spike issue! I will keep looking.

The restart batch file sounds like a great idea ED E! One question - how to I exit the loop when I need to?

ED E

Metric Tensor wrote: »

The restart batch file sounds like a great idea ED E! One question - how to I exit the loop when I need to?

In the 45 seconds you can just drag it out of the folder. Or run

shutdown -a

to cancel it once off.

Serephucus

Alright, the crackling noise isn't an unknown any more.

As for the weird temp spikes, is there a newer BIOS out for the motherboard?

Metric Tensor

There is Serephecus. I'm going to download it and install it ... should have weeks ago but didn't bother!

I think the fact that the AI Suite is misreading the temperature might relate to my problem.

dreamers75

My last H2o AIO developed a similiar problem, still to this day I have no idea what it was. There was a noise but only every other day and the CPU temp would increase again only every other day.

I just got a new AIO as i tried everything that was posted here.


Still no clue what caused it.

Cuddlesworth

Metric Tensor wrote: »

There is Serephecus. I'm going to download it and install it ... should have weeks ago but didn't bother!

I think the fact that the AI Suite is misreading the temperature might relate to my problem.

Don't think it does, there is a sensor on each core and one for the overall chip. AIsuite uses the overall sensor.

Metric Tensor

I know some there is an overall temperature reading as well as each core temp but the temperature AI suite was reporting was 20 degrees lower than the sustained 60ish that each core was sitting at during the stress test.

Metric Tensor

It did it again today lads and it took two restarts to get it back to normal. Have visitors too so I couldn't do any work/diagnosis.

Started it and the temps were floating in the 40s. Restarted and they were in the 80s and increasing up to 100ish if I did anything. The restarted a second time and sitting at 20!

Mea Culpa though - I haven't had a chance to do any of the great things suggested here. Just needed somewhere to offload and it's not a topic that the visitors would really be interested in!

I'll keep you up to date!

Metric Tensor

So - happened again this morning. Probably about 20 start ups since it happened last time.

I had used the AI suite to change pump profile to be always on a full blast but it still happened. I think the next step is to run the pump directly from the PSU.

Oh and also install the mobo update that I keep clicking later on

Metric Tensor

Right - so this happened a few more times earlier this year and then I uninstalled the ASUS AI Suite (which sometimes over-rules the motherboard) and it had not happened in ages .......

Until this morning!!

I've discovered that a full shut down and starting up again usually solves this issue but sometimes a restart does not. So I restarted and went straight to the BIOS. The temps were still high and when I went to the section that reads the fan speeds it said "N/A" next to the CPU pump header rather than a value.

I shut down and when I started up next time it gave a value rather than the "N/A" and the temperatures are back to normal.

So it seems the motherboard cannot see the pump on certain rare occasions and therefore doesn't power it up until it finds it at the next start up.

As far as I can see my options are as follows:

1. Power the pump directly from the PSU which would help diagnose if the issue I am encountering is pump or MB based.
2. Plug the pump into one of the spare fan headers on the MB and set that "fan" to run at 100% all the time.
3. Some people on line suggest connecting the pump to the PSU fan header and setting this to 100% all the time and then connecting the fans to the OPT fan header with a profile.

I'm undecided which option I'll take yet. I can see that 1 makes the most sense but for some reason my OCD would prefer 3! I also need to check if I have an adapter to allow me to run the pump directly off the PSU.

Any thoughts or ideas greatly welcomed!

Serephucus

I'd personally go with one here. It'll be the fastest way to see if you have a problem, and given how sporadic it seems to be, that's probably best.

It could be the header I guess. I'd tend to think the pump would fail before a motherboard header does, though that's only anecdotal - I've never had a motherboard fan header fail on me, but I have had a pump die.

Cuddlesworth

Serephucus wrote: »

I'd personally go with one here. It'll be the fastest way to see if you have a problem, and given how sporadic it seems to be, that's probably best.

It could be the header I guess. I'd tend to think the pump would fail before a motherboard header does, though that's only anecdotal - I've never had a motherboard fan header fail on me, but I have had a pump die.

I've got a x99 pro and it has done some weird things with the fan headers and custom fan controls. I ended up turning most of them off.

Metric Tensor

I don't think there is much possibility of the header "failing" - I think that MB will only deliver power to the header when it sees a fan/pump present.

I think the issue might be that in certain circumstances (I don't know what they are) the MB does not see anything connected to the pump header and hence sends no power.

The question is - is that a defect in the pump or just a "communication issue" between a working MB and working pump.

I think I'll try putting it straight into the PSU at lunchtime if I can find a connector handy. The problem being that it could be months before I find out if that has worked or not!

ED E

Not having read the IETF/whoever spec:

One would assume the header provides a constant voltage on the +- pins and then reads back the clocking signals from the sense pins(one pulse per revolution). So the power should be there whenever the system is in a non hiber c-state. If its not doing this there's a microcode bug or a hardware failure. You can read it with a multimeter pretty easily if you like.

Metric Tensor

Just to update this thread in case anyone finds it via a search.

I swapped the pump power supply onto the Cpu_Fan header and set it to 100% permanently and put the radiator fans onto the CPU_Opt header (with a variable profile) and I haven't had the issue since. 4ish months now and counting. Fingers crossed that this has solved the problem but I'll pop back here if the issue raises it's ugly overheating head again!!


** Going to complete stick my head in the sand with regards to the fact that the first answer to my post by ED_E said exactly this at the very start!