My i7-14700KF is thermal throttling, including while gaming, and I'm trying to determine whether the cause is CPU mounting pressure, BIOS settings, thermal paste, or a problem with the AIO cooler.
My system includes an Intel Core i7-14700KF, an ASUS TUF GAMING Z790-PLUS WIFI motherboard running BIOS 1836, an ASUS TUF GAMING LC II 360 ARGB 360mm AIO, an ASUS TUF RTX 4080 SUPER OC, and an ASUS TUF GT502 V2 case.
The CPU is still using the standard Intel retention mechanism. I have not installed the LGA1700 contact frame, replaced the cooler, or remounted the block yet. The current thermal paste is Cooler Master MasterGel Pro V2, applied about a year ago in an X pattern.
I recorded separate HWiNFO sensor logs for Cinebench, Hogwarts Legacy, Cyberpunk 2077, and God of War, resetting the statistics before each test. Some of the CPU power readings reached around 260 W, although I am not sure whether that was only a brief peak or sustained power draw.
For maintenance, I plan to clean the system, apply Thermal Grizzly Duronaut paste, install a Thermalright LGA1700-BCF Black contact frame, and remount the AIO block. I will compare temperatures and throttling behavior afterward. If the results do not improve, I may investigate the cooler further.
Before concluding that the AIO is faulty, what should I check? Does this behavior suggest a serious CPU problem, or is it more likely related to power limits, voltage, mounting, or contact? I would also appreciate guidance on safe BIOS undervolting settings if the maintenance does not resolve the issue.
3 Answers
A 14700KF can draw close to 260 W under heavy all-core workloads, so that number by itself does not prove the AIO is defective. Check whether the power is a short peak or sustained draw, and look at package temperature, effective clocks, thermal throttling flags, pump speed, radiator-fan speed, and CPU voltage in HWiNFO. Gaming temperatures can also be affected by case airflow and GPU heat.
Before blaming the cooler, remount the block carefully, verify that the pump is running at the expected speed, confirm the radiator fans are responding properly, and check that the radiator is not restricted by dust or poor airflow. A contact frame may improve LGA1700 contact, but it is not a guarantee of a dramatic temperature reduction.
If temperatures remain too high after confirming the cooler is mounted correctly, reduce CPU voltage or power limits gradually rather than jumping straight to aggressive settings. A common approach is to apply a small negative Vcore or adaptive-voltage offset, test stability, and then increase the offset in small steps only if the system remains stable.
Be careful with advice to disable protection features such as IA CEP or to force a fixed voltage limit. BIOS names and behavior vary by motherboard firmware, and disabling protections can hide instability or create unsafe operating conditions. Save a BIOS profile first, change one setting at a time, and validate each change with Cinebench, a longer stress test, and normal games. If you are new to undervolting, start with Intel default limits and a modest offset rather than copying an aggressive value.
I have not undervolted before, and the BIOS guides I found were confusing. I’ll first compare the results after cleaning, repasting, installing the contact frame, and remounting the cooler. If throttling continues, I’ll need a careful step-by-step explanation of the relevant BIOS settings.
For thermal paste, the exact pattern matters less than using an appropriate amount and getting even mounting pressure. A narrow line along the long center axis of the CPU is a reasonable method; it does not need to be excessive because the cooler pressure will spread it. Clean both surfaces thoroughly, avoid lifting the block after it touches the CPU, and tighten the mounting screws gradually in a cross pattern.
The contact frame should be installed according to its instructions and tightened evenly. Do not overtighten it, since uneven pressure can create new contact problems.
So the line should run along the long direction of the processor rather than trying to divide the chip in a particular visual orientation? I’ll use a small, even line instead of applying a large amount.
The 260 W reading may have been only a brief spike, so I’ll run another controlled test and check whether that power level is sustained. I plan to install the contact frame, replace the paste, and remount the block before trying to change voltage settings.