Jump to content
Corsair Community

Η100v2 AIO Cooler - Not efficient cooling


ggiatrakos

Recommended Posts

Hi all,

 

I am having problems with my H100v2 since the day I bought it, hope someone here has the experience to help me out.

 

My system is a brand new Intel i9 9900K CPU on a Gigabyte Z390 Aorus Master motherboard. Whenever CPU is stressed, temperatures jump to 80-90-100 degrees immediately, just after milliseconds the load starts. I don't know if this is normal to happen or if should the temperature should ramp slowly. It feels as if no cooling is installed at all.

 

The strangest thing is that CPU temperature rise does not affect the coolant temperature in a logical manner. Even after many minutes of running at high 80s / 90s, coolant temperatures raise very slowly, and never surpass 33-34 deg C. This is as if the heat is not correctly transferred to the cooler. I even tried lowering the pump speed and fan speed to the absolute minimum, hoping for coolant temps to rise, but no change at all. I keep getting a deltaT between CPU package and Cooler's liquid of 50 deg C! Please check the attached video.

 

I removed the cooler's head from the socket slot several times to see if the cooling plate touched the CPU correctly and seems everythings fine, even the paste pattern seems to perfectly match the CPU aluminum cover.

 

Thank you in advance for your help.

2019-04-30_21-26-44.zip

Link to comment
Share on other sites

  • Corsair Employee

Hello,

 

The 9900k is a very hot chip. The chip will spike quickly to the higher temps. However, the coolant not rising is a different issue. It could be a related to a couple of things. Do you have the header set for DC/12V/MaxSpeed? If so, I believe you may have a defective pump. I would advise getting a ticket open with our team and we can further assist.

Link to comment
Share on other sites

+50C coolant to CPU differential is not uncommon for a 9900K, although this value is largely determined by Vcore. You are running 1.35v and that is quite a bit for a 9900K. I would say +50C isn't so bad for value. On the bright side, I don't think you need 1.35v either. You haven't listed your clock or base Vcore settings, but I will hazard a guess you are running on full auto settings and with whatever GA uses to set all the clocks to maximum turbo speed. You should be able to take a large chunk off the top end by dialing in your BIOS settings, including specific clock speeds and Vcore levels.

 

I don't think you have a contact issue. Even with a linear load stress test, a bad mount will twitch on temps quite a bit. Your's are steady. The continued climb +1C at an interval is the increasing coolant temp. Coolant temp is the baseline CPU temp at 0 Vcore and each +1C of coolant temp is +1C to CPU temp. On a 240mm cooler, I would expect a coolant rise of +8-10C for a 9900K at 5.0 GHz. Once you got past 15 minutes or so, other factors like case design and airflow may affect the stability of the coolant delta.

 

I think for now, if you can share your BIOS settings, that may explain much of what you're seeing. Then a few revisions should cut that down.

Link to comment
Share on other sites

  • 2 weeks later...

Hi all, thank you for your answers.

 

C-attack, it seems that the Vcore values show at the iCue menu are false.

 

I have set V-core to 1,33V at 5.0Ghz from the BIOS, as lower values cause the intel stress test to fail as soon as it starts (blue screens). From 1,325V and above the CPU becomes more stable, so I left it at 1,33V. Still the test fails after 10-12 minutes.

 

 

I am runnuing a new Intel ETU stress test now at the above settings and

- HWiNFO shows:

  • Gigabyte Z390 Vcore (ITE) IT8688E: 1,356V
  • Gigabyte Z390 Vcore (ITE) IT8792E: 1,331V
  • Core #0-#7 VID: 1,290 volt (varies form 1,270-1,290 is this reasonable not to match the Vcore set from BIOS?)

- iCue shows:

  • 1,36V

 

During the 15minute test at extreme pump / extreme fan setup, the coolant temperature raised only 1,7 degrees after 10 minutes (started from ~34 and climbed up to ~35,7degrees). Please Note: the PC froze after 14 minutes.

 

I ran the test again at quiet pump / quiet fan setup and the test started from ~34 and climbed up to ~38 degrees.

 

In both cases, I can't see the 8-10 degrees difference that you mention.

CPU was running at 70C package temp.

 

Then I started the Prime 95 benchmark whith which is more power consuming.

The temps rose at 99-100C package temps immediately.

After 2minutes, coolant temp was 39C (delta-T of only 5,0C)

After 3minutes, PC crashed.

 

Also, I am not sure what you mean here:

 

Coolant temp is the baseline CPU temp at 0 Vcore and each +1C of coolant temp is +1C to CPU temp
Link to comment
Share on other sites

First, 1.356 rounds up to 1.36V. Also, vCore does swing quite a bit. A few milliseconds difference in read times between the two apps would give you different readings. Next, there's the Load Line Calibration that can allow/cause the voltage to overshoot the settings that you put in the BIOS of the CPU voltage.

 

You also need to accept the fact that the 9900K runs hot and that you may not have won the silicon lottery. You won't be able to hit the same voltages that you can with earlier chips, even with Intel's soldered TIM (which, from the reports I've seen, is only marginally better than the non soldered TIM in earlier chips but is harder to de-lid).

 

As for this:

Coolant temp is the baseline CPU temp at 0 Vcore and each +1C of coolant temp is +1C to CPU temp

Your CPU cannot be any cooler than the coolant temperature. That's the baseline. This is due to the pesky laws of thermodynamics and you won't get around it. The warmer your coolant, the warmer the CPU. Period. Again - pesky laws of thermodynamics.

 

I would suggest that you revert to stock and run the same tests again. And, TBH, 1.35V is a lot of voltage for any of Intel's recent chips. Even earlier chips would run hot with that amount of voltage applied to them.

Link to comment
Share on other sites

VID and Vcore are not the same thing. However, I would not get too bogged down in the difference. It seems you do need ~1.33 Vcore or thereabouts for your set-up. You might be able to reduce it further with LLC adjustments or other small tweaks, but that is more complex and a longer term examination. These control how much overshoot you see on the Vcore and it would be normal to see +0.03V on medium settings.

 

As for the cooler functionality, I don't really like the +1.7C in 10 minutes. Intel XTU is mild, but nothing is quite that mild. On the other hand, don't use Prime for this task. It is not effective for testing what you want to know and it is not a good predictor of daily stability either. To borrow an analogy, you are putting the CPU closer to the flame and that is not going to help establish if the cooler is working properly. It only gets you into trouble faster. Stick with Intel XTU, CPU-Z linpack or other mild tests with predictable results. The only thing predictable about Prime 95 is it will be very hot at the CPU pins.

 

I don't feel like enough heat is in the coolant stream and this usually suggests a contact problem between the CPU and pump head. The bracket in back is the most problematic. Make sure it is turned the right way and each thread from the front is going all the way through. If you lightly press down with two fingers on the pump, do you see CPU temps drop under load?

 

Heat is conducted across the cold plate both ways. Heat will leave the CPU, but it will also return from the coolant stream. Thus, the coolant temperature is the minimum possible CPU temperature. When your system is powered off and the coolant is 24C (or whatever), the CPU will also be 24C. As soon as you power on, heat is applied from the pin side of the CPU and the offset changes. Where this is relevant is in fan control. Fans and pump speeds have no effect on the interaction between the CPU pins and the CPU. That is where heat is created. It only becomes a cooler thing once it is conducted into the coolant stream. If you coolant temp has gone +4C since you started doing something, the most you can possibly reduce the CPU temperature is 4C with any fan or pump speed combination. Mostly this is a reminder to people not to run the fans at 2000 rpm. It doesn't really help and won't do anything for your Prime 95 temps. The cooler can only get rid of the 4-5C it has on its side of the cooling equation. All modern CPUs are limited at the pin side due to voltage/temperature. If not, then we could all use 6 ft tall external radiators and run 7GHz@1.7v. Obviously we can't.

Link to comment
Share on other sites

×
×
  • Create New...