10850K temps extremely high

[DarnieArnie]

Earlier today while gaming I noticed my fans ramping up, I launched HWMonitor to check out what my package temps were and they were extremely high. The highest I saw was 95c, before I've never seen higher than 70c-80c(during a Cinebench test). At idle package temps sit around 35c-45c, which has been a normal level since building this computer. Any advice would be helpful, thanks. I'll attach a screenshot I took with a game launch with temps, voltages, etc. 

 

Specs:

i9 10850K (stock clock speeds)

MSI MPG Z490 Gaming Edge WIFI

MSI MAG 360R AIO

Crucial Ballistix DDR4-3600 2x8GB

Asus Strix 1080TI

 

[DarnieArnie]

51 minutes ago, Schnoz said:

Some motherboard manufacturers impose lax power restrictions that allow the CPU to turbo for unlimited periods of time, which is probably what you're seeing here. To resolve this, disable "Multi-Core Enhancement" or something similar in the BIOS.

While trying to find MSI's equivalent to "Multi-Core Enhancement", I came across a Reddit post of someone with similar issues. His was too high of voltage, and he found that MSI Dragon Center was boosting his CPU. Could my issue possibly be with Dragon Center? Also, at idle I'm clocking at 4.8-9GHz, should I adjust that in anyway? Stock clock is 3.6GHz per a Google search.  EDIT: Still spiking during gameplay with Enhanced Turbo disabled, 87c was max before I closed the game.

[DarnieArnie]

11 hours ago, Schnoz said:

You could definitely try uninstalling or disabling Dragon Center, though I don't have any experience with that application. Though, I'd imagine that you could have better results with undervolting the CPU in BIOS settings.

 

The idle clock shouldn't be a concern; the CPU isn't spitting out much heat at the moment.

Any advice on undervolting? I've never messed with any of that before. What values should I try first? To my eyes the voltages it's getting under load are around the stock numbers, so I'm just confused as to why I'm suddenly experiencing these drastic spikes in temperatures. Should I reapply thermal paste to the CPU/block? 

[DarnieArnie]

4 minutes ago, Schnoz said:

Reapplying thermal paste first is a good option; I recommend Thermal Grizzly Kryonaut. If temps still don't work out after that, then I'll tell you more about undervolting.

On hand currently I have access to ArticSilver and Thermal Grizzly Conductonaut, I ordered the Conductonaut for this build but got kinda sketched out after reading more about it. I couldn't find anywhere saying the block on the AIO was in fact copper (I'm 99% sure it is) but didn't want to risk anything given the current market on computer components. 

[DarnieArnie]

4 hours ago, Schnoz said:

Arctic Silver is...okay. I wouldn't recommend it for this application, especially since it apparently has a tested thermal conductivity of 0.9 W/mK, which is well below average.

 

Conductonaut, I wouldn't recommend--it leaves an unremovable coating on surfaces and touches, and it will short out contacts if it drips. Also, it's super hard to apply and can get messy.

 

To undervolt your CPU, lower the voltage by 50 mV in the BIOS, and run Prime95 on it until temperatures level out to check stability. Keep lowering the voltage until Prime95 becomes unstable, then undo the last undervolt and run Prime95 on it for like an hour to verify that it's stable.

Called around to several stores, and none carry Thermal Grizzly. Got some ordered but it will not be here until next week. Might tinker with the undervolting and see if that shows any results. Thanks for the help, I'll post an update when I can. 

[DarnieArnie]

19 hours ago, Schnoz said:

No problem! If you could consider marking the reply you found most useful to you as the solution so others with a similar problem could easily know what to do, that would be greatly appreciated.

 

Undervolting works like magic on some processors. Even a relatively conservative 100 mV undervolt on my 5700 XT cuts its power draw in half and ironically results in higher performance, as it boosts without hitting its power limit. Though, this is a completely different architecture and manufacturing process, so mileage will vary.

Okay..so this is what my BIOS is showing voltages to be set to. The second picture is voltages at idle. Where would you start if at all to undervolt? 

[DarnieArnie]

1 minute ago, Schnoz said:

Start by the CPU Core Voltage section, and lower it each time by 0.05V. Run Prime95 after each voltage-lowering for around 5 minutes to check its stability. You'll know you've hit the limit when trying to run Prime95 results in a crash; in that case, undo the last voltage-lowering and run Prime95 again for around an hour to check for continued stability.

I found another post with similar issues, and they mentioned using ThrottleStop and doing offsets with it and using its test bench. Well, once I get to an offset of more than -35mV Windows will blue screen. And the temps are still insanely high. So I'm wondering if there could be a problem with the AIO. HWMonitor is showing RPM on the pump, and there is slight vibration in the pump. But i'm not sure, kind of stumped at this point. 

[DarnieArnie]

4 minutes ago, Schnoz said:

On most good pumps, the vibration should be basically zero. How is your AIO mounted? An air bubble may have lodged itself in an inconvenient location, causing the bad temps. Would you mind sending a photo?

 

[DarnieArnie]

13 minutes ago, Schnoz said:

On most good pumps, the vibration should be basically zero. How is your AIO mounted? An air bubble may have lodged itself in an inconvenient location, causing the bad temps. Would you mind sending a photo?

This is the other post I was referencing.

[DarnieArnie]

10 minutes ago, Schnoz said:

An air bubble might have gotten stuck in your AIO pump. Try flipping the waterblock around and see if it fixes anything. If that doesn't help, try underclocking the CPU by around 200 MHz.

 

Also, it's odd that only a 35 mV undervolt will result in a BSOD. Maybe your particular unit just barely made the 10850K bin.

I took the block off last night and went ahead with applying the Artic Silver paste, which did result in slightly lower idle temps from the factory paste that came on the AIO. Should I turn off all Intel turbos in the BIOS? In the BIOS it shows the clock speed set to 3.60GHz, but it's sitting at 4.8-5.0GHz constantly across all cores at idle. 

[DarnieArnie]

3 minutes ago, Schnoz said:

Repasting it with Kryonaut will help even more, as its thermal conductivity is 12.8 W/mK, significantly higher than the 0.9 W/mK reported by Arctic Silver. It's also a bit less viscous, allowing it to spread more thinly. Furthermore, Arctic Silver 5 needs around 200 heat-cool cycles before reaching maximum potential, according to Arctic Silver. When you removed the cooler, did you notice that the paste was spread in a thick layer? That could also be contributing to poor temperatures.

 

I don't think it's the best option to turn off Turbo Boost entirely, as that can significantly reduce performance. High idle clocks are fine--you're on a desktop so it's fine with drawing a bit more power at idle, and it's barely doing anything so it's not spitting out much heat. I'd try simply lowering the boost clock by around 200 MHz, since that can drop power draw significantly (power draw increases exponentially with clockspeed).

 

I'm going to reapply the Kryonaut when it comes in next week. And when I took the cooler off it was spread more to one side than the other, had uneven pressure on the mounts. 

 

Also, I turned off Intel Turbo, and set the clock speed ratio to 42, 4.2GHz, and temps are holding steady at much lower levels. Around 60C, with a max of 74C. 

[DarnieArnie]

On 1/8/2022 at 7:14 PM, Schnoz said:

Looks like mounting pressure could be part of the issue then. I've definitely noticed that some IHSes can have significant variations in flatness, but it could also be the clamps. Let's see if the Kryonaut fixes it. If it does, please mark the post mentioning Kryonaut as the solution through the button on the post, so others with the same problem will be able to quickly find a good fix.

 

By the way, if Kryonaut doesn't work as well as you'd like it to, they also sell Kryonaut Extreme, which is more expensive than ordinary Kryonaut but has a thermal conductivity of 14 W/mK, the highest conductivity of any non-metal paste I've ever seen. It's also bright pink, so that may or may not be a bonus depending on how you see it.

 

Also, great to hear that temperatures are now under control! Since you have such a large thermal margin, you can try raising the all-core clock to 4.5 GHz. I imagine that'd get you to around 80 C, which is still a safe temperature.

So my issue was my cooler being dead. Turned my computer on and heard some terrible sounds from the cooler. Ordered an Arctic Liquid Freezer II to replace the MSI cooler. Seems to have been a very common issue with that specific cooler, good to know after the fact, smh. Applied the Kryonaut with the new cooler and temps are looking great now. Thanks for the assistance you offered, will mark thread as solved.