Corsair H100x Running CPU at 100'C Underload

[SharpShot.SCL]

Hi, I have just installed a new Corsair H100x into my new build and instantly I have been having issues with CPU temps. Running everything at stock and no over clocks applied. AIO block plugged into SATA power and AIO_PUMP, fans on Y splitter plugged into CPU_FAN

 

I wanted to test performance of my new AIO and ran Cinebench R23, instantly the load ran upto 100% on the CPU using HWMonitor but the temps also matched that, sitting in the high 90s and hitting 100'C. This caused me concern as I upgraded from a 7y/o Corsair H60 AIO to allow sufficient cooling to allow an over clock to be applied. Running the same benchmark with the H60 installed, the CPU barely reaches 80'C on HWMonitor. I have thought it could have been pump failure but the CPU will cool down to 30-40'C straightaway after running R23. That also said I did hear water moving around the H100x the first time I started up the system with it installed.

 

So far I have tried reapplying the pre applied thermal compound with MX-4 and no success. Also checked for any plastic or other packaging and have found nothing. The radiator and pipes do not seem to heat up when the CPU reaches these temps but the block is warm-hot to the touch, which based on other posted lead me to think of possible pump failure

 

If anyone has had similar issues or knows of anything else I can do to test or further set up my AIO to get the temps down please let me know.

 

Thanks in advance.

 

TLDR: AIO cooled CPU reaching 100’C under full load but will cool to approx 30’C after. Old AIO never went above 80’C in the same conditions.

Edited January 19, 2021 by SharpShot.SCL
Title Typo

[c-attack]

Pump failure - You have 60 seconds before at idle before the entire system shuts down. You likely are unable to boot up after.

 

Partial flow blockage or other flow issue - Load test starts off OK, but the CPU temps starts an immediate phased step-up after the initial jump to 70-80C. For example, you start R23 and the CPU temp hits 72C instantly, then gains +1C ever 3-4 seconds until it hits 100C. That is coolant temp rise and it also raises the baseline CPU temperature. You might also see the CPU temp slowly increase at the desktop level with a slow march up in temp from the moment you boot up. Regular desktop actions do not cause alarmingly high temps, but as the baseline CPU temp increases, so will the higher marks when opening programs.

 

Contact issue - You start R23 and it hits 80-90C immediately. Small a nd slow 5-8C increase over time if you manage to keep it running. You also will see this on the desktop with CPU temp spikes to 60-70C when opening browsers or other programs. When the voltage drops, so do temps but they instantly hit high marks any time the voltage is raised.

[SharpShot.SCL]

Thanks for the quick reply, it sounds really likely to be a flow or blockage issue. When I run R23 it will run up to 70-80C then slowly climb up to 90-100C before the test completes. Once completed it will return steady to 30-40C.

 

Is there anything I can do to fix that myself or is it going to need an RMA?

[c-attack]

That seems like some type of issue where the cooler cannot get rid of its heat. Flow restriction is one possibility, but not the only one.

 

1) Check for anything that might prevent heat from being expelled from the radiator. Examples might be a dust filter plastered against one side or the other, an environmental restriction like the case in a cabinet and that basically becomes a hot box, or other case peculiarities. You appear to have a 4000D, so I have no idea what that would be and don't see that as a possibility.

 

2) Run a different test to confirm the behavior. It is natural for Cinebench to get warmer on the CPU as the render progresses, in part because of latent heat retention. Try using the stress test contained with CPU-Z (bench tab). It is mild and completely linear which makes it a good test tool for those will potential cooling issues. When you start, 1 second later it should hit 70C (or whatever) and hold. It will increase +1C for each +1C in coolant temp increase. That is a value you can't see, but we can predict it the increase. I would expect +6-8C for 240mm radiator and a 10700K. However, there is a large caveat in #3.

 

3) Really wide range of power (watts) possibilities on Comet Lake. If you have Asus MCE enabled in the BIOS, that removes all the Intel power limits. R23 is a 10 min loop test and you could potentially draw 230-250W from an 8 core at 5.0x10. I am going to run mine in second, but 250W is getting to bit a bit much for a 240mm without seriously prioritizing its function (high fan speeds, thick radiator, etc).

 

**Just ran a few laps of R23 on my 10900K. HWiNFO puts the wattage at 240W and my other monitor puts it at 265W. I am running 5.2x10 with all power limits disabled, so this is a worst case scenario. In an overclocked state, with 2 less cores I would expect you to be in the 200-215W range. That corresponds into approximately a +8C rise on a 240mm with the fans at 1300 rpm or so. 240mm radiators don't scale well with the fans below 1000 rpm, so you could easily see +20C at 700-900 rpm.

Edited January 19, 2021 by c-attack

[SharpShot.SCL]

I have tried a few different things; I currently have the radiator mounted on the front of the case, fans pushing air through it with the tubes at the bottom. I have also tried the tubes at the top and that actually made it slightly worse but could have been within margin of error 2-7C warmer.

 

1) I am using the 4000D Airflow and there are no restrictions on to radiator or fans at all, I have also run it with both side panels and front panel and dust filter off.

 

2) I have now run CPU Z stress and the CPU did instantly climb to 70C and climb up to between 78-82C while that was running and again cooled back down to around 30C after.

 

3) Power is something I am thinking about, the RM850x is new, previously ran CX600. Asus MCE is not enabled to my knowledge, I will check again. Power mode is set to auto in the BIOS and limits are in place, again I will check to confirm. Power draw does not seem to be that high in HWMonitor and the fans are all running well over 1000rpm under load.

 

I have attached photos of the layout inside the case, including my change to pipes up, along with HWMonitor display while running CPU Z. The max temps reached in those pics are from the R23 run.

 

 

[c-attack]

You have a 21C difference between hottest and coolest core on the max. That is generally out of bounds for any 10000 series CPU I have seen and generally suggests a partial contact issue. I would take the block off and see if the TIM is spread evenly and also check the mounting post depth through the MB and backplate. When this happens, you are just a little off or have uneven pressure. Make sure you gradually tighten the thumbscrews a little at a time, working around or across from post to post.

[SharpShot.SCL]

Seems like this has solved my issue! I had been using my old back plate from the H60 and standoff screws. Now I have used the included screws and fixed the back plate with the pre-applied adhesive strip and re-applied MX-4 and the results from R23 are vastly improved, now under full load sitting around 64C with only an 8C difference from the hottest and coolest cores.

 

There is a similar story with CPU Z stress, ramping up to around 50C and not even reaching 60C.

 

Hopefully this is the issue resolved now, thank you very much for your advice c-attack!