Question Brand new i9-12900K + Waterforce X240 jumps to 100ºC immediately after stress test starts. Thermal throttles the entire time.

Mar 24, 2022
I built a new system with all brand new parts, and when I perform a CPU stress test the temps go to 100ºC immediately, like seconds after CPU Usage goes to 100%, and then just stay there for the entire test with Intel XTU giving me the "Thermal Throttling" warnings. Cinebench R23 30min Stability Test completes just fine and gives me a decent(?) score of 26408 pts, it just thermal throttles for the entirety of the test.

The first obvious thing to do was checking the thermal paste and pump mount, so I took it out, verified that the thermal paste was evenly distributed (it was, CPU perfectly covered) and then re-applied it also making sure to really tighten the screws this time. In doing so, I realized I hadn't tightened them as much the first time, so I was feeling quite confident that that was the issue... but nope. Once again, even after a cold boot as soon as CPU usage goes to 100% the temps jump to 100ºC and begins thermal throttling.

Here is a CSV export of all sensors from Intel XTU during the 30min Cinebench test:
Look at how it jumps from idle 28ºC to 100ºC in like 7 seconds: View:

  • Re-applied thermal paste, and REALLY tightened the screws
  • Thermal paste was evenly distributed when I took it out the first time.
  • Idle temp 27~31ºC
  • It's a cold day ~14ºC
  • Brand new system (0 dust)
  • Fully tested other components (RAM, SSD, GPU), no issues
  • Manually set Pump and Fans speed to MAX
  • No OC other than enabling Memory's XMP profile (same behaviour with XMP disabled)
  • Not just Cinebench, same behaviour on Intel XTU stress test
  • Cinebench R23 MultiCore 30-minutes test: 26408 pts
  • Top-mounted radiator: View:
Full parts list
  • CPU: Intel Core i9-12900K 3.2 GHz 16-Core Processor
  • CPU Cooler: Gigabyte AORUS WATERFORCE X 240 60.07 CFM Liquid CPU Cooler
  • Motherboard: MSI PRO Z690-A WIFI DDR4 ATX LGA1700 Motherboard
  • Memory: Kingston FURY Renegade RGB DDR4-3600 32GB (2x16GB) CL18
  • Memory: Kingston FURY Renegade RGB DDR4-3600 32GB (2x16GB) CL18
  • Storage: Western Digital Black SN850 2 TB M.2-2280 NVME Solid State Drive
  • Video Card: Zotac GeForce RTX 3070 Ti 8 GB GAMING AMP Holo Video Card
  • Case: Fractal Design Meshify 2 Compact TG Light Tint ATX Mid Tower Case
  • Power Supply: Corsair RMx 850 W 80+ Gold Certified Fully Modular ATX Power Supply
Mar 24, 2022
Where is the pump connected on motherboard?
that's an interesting question, what's your hypothesis?

the only connection to the motherboard is through the USB 2.0 Cable. Power comes from a SATA connector, but that's directly from the PSU.
There are no other left-over connectors, and that's what the manual instructs you to do, so I'm not sure what I could have done incorrectly there.
Also, the AIO Driver correctly reports both pump and fans RMPs, and I was able to set them to MAX during testing

Nice PSU cables.
12900 is a monster. 240mm rad might just not be enough for it.
Hey thanks! they are the stock ones, nothing fancy.

Yeah, that's what I'm getting from my troubleshooting. I researched it beforehand and people were saying that it should be close but fine. I can't fit a 360 in this case, so I may have to take the performance hit (and "wasted money"). That's unfortunate.


Normal for that cpu, if you do nothing to it.
Check some 12900K thermal threads on TechPowerUp, and Linus Tech Tips', Kitguru, and Hardware Unboxed's reviews. Typical result on non-gaming applications.
That cpu needs to be manually tuned if you want it in your comfort zone. Probably the easiest step you can do is to turn down Maximum Turbo Power - and Processor Base Power* - until you reach a temperature range you are more comfortable with.
[On some boards, MTP and PBP are the same, instead of PBP being lower than MTP.
Mar 24, 2022
Thank you everyone for you prompt help and insight, I really appreciate it.

Seems like the bottom line here is that I screwed up a bit and went with a very hot chip without sufficient cooling.
I'll explore the options of undervolting or potentially upgrade to a 280 AIO and maybe add an extra fan or two to the case.

Overall I'm not too worried as the system is stable, and I won't realistically be pushing it to its limits, I'm just a bit bummed out that I made a mistake on my purchases but all in all I just lost a bit of potential performance and money I guess... the world keeps spinning.

Thanks again everyone!
If you are uncomfortable with current temps, you can consider lowering target all core clocks by 100-200 MHz and/or trying a very small core voltage reduction, as either/both will help with produced CPU heat. (Intel's XTU will allow the manipulation of P/E core target clocks, boost duration, core voltage offset, AVX offset, where slightly lower clock speeds can be selected just for AVX workloads which do produce more heat, etc...)
Don't touch voltages and turbo clocks, the only thing you need to change is the max power the system should draw.
Reducing voltages can make the system be less stable or crash when running something weird and lowering clocks outright cuts down your performance everywhere.
Change these two values, turbo boost short and max, these could be at ~4000W right now because mobo makers put this at max, benches have shown that anything above around 240W doesn't give you any more performance while increasing heat by a lot.