CPU Thermal Throttling

[Jeeby]

So I just finished my 3rd build, running an i9 13900K on a ROG Maximus z790 Hero cooled by a Corsair iCUE H150i Elite Capellix XT. I installed the cooler using the pre-applied thermal paste, I have the pump cable plugged into the CPU_FAN header and the 3 fans plugged into the iCUE Commander. I have the three fans running at 2150 RPM, the pump at 2600 RPM and the CPU_FAN header running at 'Full Speed' in BIOS. Upon running an AIDA64 test, I get up to 6% throttling, and the cores are reaching 85-100+ under load. The pump's liquid stays in the mid 30's for temp.

 

Everything is out of the box and the only BIOS setting changed is that CPU_FAN header. Any help would be appriecated! Thanks.

 

Picture of test: https://imgur.com/a/hxyI24R

[Godlygamer23]

Since one core is significantly cooler than the other, by around 30 degrees C, I would check the mount to make sure it's even across the CPU.

Edited August 8, 2023 by Godlygamer23
Minor correction on temperature value.

[Jeeby]

Yeah I need to wait for a new thing of Kryonaut first, I'm not sure where my old one went when I delid my old CPU. Don't want to dismount the pump yet until then, I figured the cores not matching in temps may have something to do with the spread of thermal paste. I'll have to check it out on the 10th.

 

I've been on a couple other forums, someone said to daisy chain the 3 radiator fans to the CPU_FAN header and to put the PWM for the pump onto the AIO_PUMP header instead. And someone else said it may be a voltage issue. I've never had to daisy chain PWM's onto one header so I'm not sure how to go about that. As for putting the pump on the AIO_PUMP header, I still need to wait for that Kryonaut since I was dumb and put the pump where the cable wouldn't reach the AIO header, so I need to dismount it anyways to turn it 90 degrees lol.

[Godlygamer23]

17 minutes ago, Jeeby said:

Yeah I need to wait for a new thing of Kryonaut first, I'm not sure where my old one went when I delid my old CPU. Don't want to dismount the pump yet until then, I figured the cores not matching in temps may have something to do with the spread of thermal paste. I'll have to check it out on the 10th.

 

I've been on a couple other forums, someone said to daisy chain the 3 radiator fans to the CPU_FAN header and to put the PWM for the pump onto the AIO_PUMP header instead. And someone else said it may be a voltage issue. I've never had to daisy chain PWM's onto one header so I'm not sure how to go about that. As for putting the pump on the AIO_PUMP header, I still need to wait for that Kryonaut since I was dumb and put the pump where the cable wouldn't reach the AIO header, so I need to dismount it anyways to turn it 90 degrees lol.

If it were related to the signaling or power for the pump, or how the fans are getting a signal, then the pump liquid temperature would reflect that. The liquid inside the pump(which is most likely propylene glycol) can take a lot of abuse. If it couldn't dissipate the heat, I would expect the liquid temperature to climb. So if it's instantly thermal throttling, I would definitely check the mount at least while waiting for thermal paste. 

 

What are the temperatures of your other cores? Also what is the actual RPM the pump is currently running at?

Edited August 8, 2023 by Godlygamer23

[Jeeby]

I don't have the PC plugged in atm, currently using my old PC. But the pump was running at 2600 RPM during the test. And during the test the cores would fluctuate between 65-100. Idle, the CPU sits at about 45.

 

Also someone in tomshardware says that if I plug the AIO pump into the AIO header, it should allow me to change from air to water, but doesn't the pump only do water?

[Jeeby]

Think I found the problem, but how did this happen? I installed the pump the way I always do, put it straight down, held it firm while i tightened the thumb screws.

 

[Godlygamer23]

8 hours ago, Jeeby said:

Also someone in tomshardware says that if I plug the AIO pump into the AIO header, it should allow me to change from air to water, but doesn't the pump only do water?

Nothing about this comment makes any sense. The header doesn't change the physical properties of the part plugged into it. 

[unclewebb]

@Jeeby

 

Consider buying a contact frame. It will buy you a few more degrees of temperature headroom before thermal throttling is necessary. Also consider reducing voltage for the core and the cache by about -50 mV for the same reason. More temperature headroom.

 

The 13900K is tough / impossible to cool out of the box during any unrealistic full load stress test like AIDA64 or Prime95 Small FFTs. 

 

 

[Whatisthis]

11 hours ago, Jeeby said:

Think I found the problem, but how did this happen? I installed the pump the way I always do, put it straight down, held it firm while i tightened the thumb screws.

 

Paste wasn’t applied over the whole contact area from the looks of it. 
 

I second @unclewebb.

[unclewebb]

9 hours ago, Whatisthis said:

Paste wasn’t applied over the whole contact area from the looks of it.

I think this is a sign that the IHS heat spreader over top of the cores is not completely flat. The crappy OEM mounting hardware that Intel uses is the most likely problem. It should be replaced with a contact frame if running full load stress tests without throttling is important. 

 

 

[Whatisthis]

23 minutes ago, unclewebb said:

I think this is a sign that the IHS head spreader over top of the cores is not completely flat. The crappy OEM mounting hardware that Intel uses is the most likely problem. It should be replaced with a contact frame if running full load stress tests without throttling is important. 

 

 

The low point is in the middle - his outside edge on the right in the pic is completely dry which is a paste application issue.

 

I do agree there seems to be some bend as well. Shame on Intel for the crap design. All they needed to do was put 2 tabs per side toward the corners…

[Jeeby]

Thanks for the info guys, I'll make sure to buy this and see how it goes!

 

Also, looking it up it looks like Grizzly makes a 12th gen contact frame on Amazon, would it matter using that on a 13th gen CPU?

[unclewebb]

6 hours ago, Jeeby said:

13th gen CPU

12th and 13th Gen contact frames should be exactly the same.

[Whatisthis]

6 hours ago, Jeeby said:

Thanks for the info guys, I'll make sure to buy this and see how it goes!

 

Also, looking it up it looks like Grizzly makes a 12th gen contact frame on Amazon, would it matter using that on a 13th gen CPU?

Get the thermalright one. It’s a much easier install. And cheaper.

[Jeeby]

Yeah seems the thermalright one is a bit more popular too.

[Jeeby]

So I installed the contact plate and it seemed to help a little bit, but it still has minor throttling at 100%. I'm curious, should I just let this go? I rarely ever see my CPU loads go even above 90% in most situations, let alone 100. Would I notice any issues in gaming/streaming etc?

 

Seems to have made it about 5-7 degrees cooler.

[Jeeby]

I ran a benchmark test on Cyberpunk and got an average 127 FPS on Raytracing Ultra settings, AIDA64 showed that the CPU stayed around 50-60C during the test, around 50% load. One thing I'm curious about since I'm not much of an expert when it comes to power management. The CPU seems to spike over 200W, even idling. Is that normal?