i9-9900k + x72 Kraken Thermal Throttle Issue

[RagingTrojan]

For a while now my stock i9-9900k, being cooled by the Kraken x72 has been thermal throttling. This usually happens when under high loads which, since I do a lot of work in Unreal Engine is a big issue. The temperature typically spikes to the late 80's when running Prime95 and occasionally ticking into the early 90s. This is all on stock speeds with default BIOS settings.

I have tried adjusting the CPU core frequency and offset settings to no avail. Below is a screenshot of the results during thermal throttling while running Prime95. Bouncing between 800MHz and 4.7ghz. I have also tried re-seating the AIO cooler several times and cleaning all the associated parts etc. 

 

[Hairless Monkey Boy]

The max temps indicate that you are not thermal throttling, which occurs at 100C. That isn't to say that your temperatures are not a bit toasty. It is a hot chip, but if I were you, I would try reseating the cooler with fresh thermal compound (maybe buy something of higher quality). Also check your fan speeds/fan curve and your pump speed.

[RagingTrojan]

9 minutes ago, HairlessMonkeyBoy said:

The max temps indicate that you are not thermal throttling, which occurs at 100C. That isn't to say that your temperatures are not a bit toasty. It is a hot chip, but if I were you, I would try reseating the cooler with fresh thermal compound (maybe buy something of higher quality). Also check your fan speeds/fan curve and your pump speed.

The speed starts bouncing down to 800mhz when under load for a little while once temps reach late 80s (I know throttling shouldnt start at this point but it seems to). I have used two different thermal compounts, Arctic Silver 5 and MX-4 and reseated the CPU cooler 3 times in the last 6 months since I noticed it happening. I have bumped up the Fan and Pump curves, even setting them to 100% at all times with no luck. 

[Helpful Tech Witch]

can you hear the pump running?

[RagingTrojan]

1 minute ago, HelpfulTechWizard said:

can you hear the pump running?

Pretty sure I can and the nzxt CAM is showing 2600-2800 RPM on the pump. 

[Helpful Tech Witch]

Just now, RagingTrojan said:

Pretty sure I can and the nzxt CAM is showing 2600-2800 RPM on the pump. 

a broken pump could also give bad data. Could you RMA it?

[unclewebb]

Looks like power limit throttling, not thermal throttling. Use HWiNFO or ThrottleStop - Limit Reasons to properly report what is going on.

 

It could also be a sensor on your motherboard sending a tbrottling signal via the BD PROCHOT signal path. ThrottleStop can report this type of throttling. Other monitoring software cannot.

[RagingTrojan]

10 minutes ago, unclewebb said:

Looks like power limit throttling, not thermal throttling. Use HWiNFO or ThrottleStop - Limit Reasons to properly report what is going on.

 

It could also be a sensor on your motherboard sending a tbrottling signal via the BD PROCHOT signal path. ThrottleStop can report this type of throttling. Other monitoring software cannot not.

So I've downloaded it and ran prime 95 as usual. It does ping the BD PROCHOT signal path. What advice can you give? 

[unclewebb]

Motherboard manufacturers do not publicly document what sensors can send throttling signals to the CPU via the BD PROCHOT signal path. It can be temperature or power related. If you clear the BD PROCHOT box in ThrottleStop, this will tell the CPU to ignore any external throttling signals.

 

Is this safe to do? No idea. The voltage regulators have their own signals for throttling purposes. Not sure what is triggering this throttling on your motherboard.

 

Why not post a screenshot of Limit Reasons so other users can learn about this issue? What motherboard do you have? Try pointing a fan towards the voltage regulator area on the motherboard.

[RagingTrojan]

[unclewebb]

When a signal is sent via BD PROCHOT, this tricks the CPU to go into full thermal throttling mode. The CPU multiplier will drop to 8 and the CPU will drop down to 800 MHz. When a sensor has gone bad, this can happen when the CPU is idle. Generally this is more of a laptop issue but it does happen to some desktop boards.

 

Do you need to run Prime95 at 4.7 GHz or greater? This app sucks a lot of power. Most apps are not going to put this kind of load on the CPU so for everything else, your CPU should be able to run at full speed without triggering BD PROCHOT throttling.

 

If you use ThrottleStop to disable BD PROCHOT, your CPU should maintain full speed but I cannot guarantee if this will be safe. The manufacturer might have put this limit in place to protect some component on the board.

[RagingTrojan]

For the most part I don't notice it. But as mentioned when working in Unreal Engine I am often hitting this limit, especially when building/rendering levels and considering this rig is built with that purpose in mind its very annoying. I'm going to try and switch out for a corsair h100 I have spare tomorrow and see if it makes a difference. 

 

I have tried lowering the clock speeds and I cab run it at <= 4.2ghz without going over 70c however still occasionally triggering the throttle. 

[unclewebb]

The CPU temperature is not the problem. It is a sensor on your board that is triggering this.

 

It could be that your motherboard is not capable of fully powering your CPU when running loads like Prime95. It could just be a defect in your individual board or it could be a limitation of all Gigabyte boards of the same model. Maybe it is a bad batch of boards where they sourced some insignificant component from a different supplier than usual.

 

It is one of those obscure problems where it can be difficult to get a manufacturer to take responsibility for. Hopefully you can find a solution. It would be nice if they replaced your board. 

[RagingTrojan]

So I threw in the h100 and can confirm you're right about it not being a temp thing, had to hold the block against the CPU and was idling at 90C but never throttled. Only thing left is as you say, the motherboard. 

[Hairless Monkey Boy]

18 hours ago, unclewebb said:

What motherboard do you have?

@RagingTrojanI don't think you ever answered this question.

[RagingTrojan]

On 11/23/2020 at 8:13 PM, HairlessMonkeyBoy said:

@RagingTrojanI don't think you ever answered this question.

 

On 11/23/2020 at 2:46 AM, unclewebb said:

The CPU temperature is not the problem. It is a sensor on your board that is triggering this.

 

It could be that your motherboard is not capable of fully powering your CPU when running loads like Prime95. It could just be a defect in your individual board or it could be a limitation of all Gigabyte boards of the same model. Maybe it is a bad batch of boards where they sourced some insignificant component from a different supplier than usual.

 

It is one of those obscure problems where it can be difficult to get a manufacturer to take responsibility for. Hopefully you can find a solution. It would be nice if they replaced your board. 

 

On 11/23/2020 at 1:32 AM, HelpfulTechWizard said:

a broken pump could also give bad data. Could you RMA it?

Apologies for the delay in responding, work and Uni etc got me busy this week.

I ordered a new motherboard on Monday, an AORUS Z390 Ultra, to replace my old Gigabyte Z370HD3. Remarkably I have seemed to hit two birds with one stone. With the new motherboard in place all temperature issues have since been resolved, with the CPU now running Prime95 non-stop for a long period of time without any throttling, remaining steady in the mid 70C @ 4.7ghz.

I specifically chose the Z390 motherboard for its 3 M.2 slots (an upgrade from the 1 on my old board) and its apparently very good and reliable CPU power solution and VRM for the price tag of £240. 

Id like to thank all of you for your input, deffo pointed me in the right direction as this time last week I was convinced it was a cpu cooler issue.