I have been running an i9-13900K in an Aorus Z790 Elite AX Rev. 1.1 for about two years with a mild undervolt: -0.075 V on the core, -0.050 V on the ring, low LLC, PL1 and PL2 at 253 W, and a 307 A current limit. With a Corsair H150 RGB, Cinebench R23 previously reached around 88–92°C, with the P-cores at 5.3 GHz and E-cores at 4.1 GHz.
I installed a Thermalright LGA1700 contact frame hoping to reduce temperatures. During the first Cinebench run, temperatures initially looked normal, but after a while the CPU began reaching 100°C and throttling the P-cores to around 5.1 GHz or lower.
Since then, I have replaced the thermal paste, tried both the contact frame and the stock retention bracket, repeatedly remounted the AIO, and tightened everything evenly in a cross pattern without overtightening. I have probably remounted the cooler around 15 times. I also heard bubbling from the pump at high speed, so I shook and repositioned the radiator and pump; the noise stopped, but temperatures remained unchanged.
The temperatures seem to start normally and then climb significantly under sustained load. Could the AIO pump, radiator, fans, or a blockage be failing? Is it possible that years under the stock bracket permanently affected the CPU or heat spreader contact, or is there another troubleshooting step I should try?
4 Answers
Air in the pump is still plausible, especially since you heard bubbling after moving the AIO. Run the pump at a fixed high speed, make sure the radiator is positioned so air cannot collect in the pump, and check whether the hoses feel noticeably different in temperature. However, after all those remounts, I would stop changing the paste and isolate the cooler with a different one.
Poor cold-plate contact usually causes high temperatures immediately, along with more dramatic temperature spikes and unusually high idle temperatures. Since the system initially behaves normally and then loses thermal performance under load, that points more toward the cooler failing to remove heat than toward the CPU suddenly changing shape. A temporary test with a stock or borrowed cooler should confirm this quickly.
Check the simple variables too: confirm every radiator fan is spinning, make sure the pump is powered from the correct header, and monitor pump RPM while Cinebench is running. Also check case airflow and whether the radiator is becoming heat-soaked. If the pump speed drops or the radiator stays relatively cool while the CPU reaches 100°C, the AIO likely has a circulation problem.
The most useful test now is a completely different cooler. If another known-good air cooler or AIO keeps the CPU under control, your current H150 is the problem. A cooler that starts at normal temperatures and then gets much worse during a sustained load can have a weak pump, trapped air, internal blockage, or poor liquid circulation. Also verify that the radiator fans keep running and that the pump RPM remains stable.
The CPU or solder could theoretically have been affected by years of bending from the original retention mechanism, but that is much less likely than an AIO problem. I would only investigate the processor or motherboard after testing another cooler.