Question 12900K Temps Going Crazy ?

[Frestyl]

I've been struggling through this issue for a few days now.

When I run any kind of test or benchmark on my 12900K, the CPU temp jumps to 100C within 1-2 seconds from a base of 25C. I'm running on a 360 AIO. I found some improvement by backing off the XMP profile on the ram and also by adding a -0.06V adjustment to the core voltage. Otherwise I have left the CPU alone.
If I was hitting 100C on sustained benchmarking I would consider bad cooling but 1-2 seconds isn't even long enough to saturate the plate, let alone remove. I've reapplied paste a couple of times.

MSI Z690 Carbon
2x16GB DDR5 6000MHz(5200 currently)
2060s OC 8G
Thermaltake 360AIO
Samsung 2TB NVME
2 other old SSDs and 2 HDDs

TIA

 

[Newtonius]

Have you checked if your AIO is making flush contact with your CPU IHS? The 12900K IHS is known for warping after a bit, like curving, making contact with the CPU cooler poor. You can test this by 1. Visually inspecting it with your bare eye, 2. Putting the CPU IHS on a flat surface like a table and spinning it, and if it spins a lot (more than 2-3 seconds) then you got a curved IHS.
If everything else you've done is correct; proper paste application, cooler is screwed in tight, AIO pump and working properly, etc, then that would be my last guess. Also what voltages are your CPU reaching when it reaches those temps, and would you say whatever that wattage may be is correct?

 

[TerryLaze]

If you have adaptive boost enabled, which it is by default, it will will always try to reach 100 to get the best performance.
Disable this if you don't want it and put in a manual all core.
https://www.intel.la/content/www/xl...intel-technologies-boost-cpu-performance.html

Intel® Adaptive Boost Technology

Opportunistically increases all-core turbo frequency when current, power, and thermal headroom exists. Works below a temperature limit of 100°C.

 

[Frestyl]

Have you checked if your AIO is making flush contact with your CPU IHS? The 12900K IHS is known for warping after a bit, like curving, making contact with the CPU cooler poor. You can test this by 1. Visually inspecting it with your bare eye, 2. Putting the CPU IHS on a flat surface like a table and spinning it, and if it spins a lot (more than 2-3 seconds) then you got a curved IHS.
If everything else you've done is correct; proper paste application, cooler is screwed in tight, AIO pump and working properly, etc, then that would be my last guess. Also what voltages are your CPU reaching when it reaches those temps, and would you say whatever that wattage may be is correct?

Thanks
I'll check that out.
Voltages are hitting 1.36 during load and drop to 0.65 when idling. Top speed is 4987MHz, lowest 498MHz, on PCores that is.

Changes I've made have reduced Power usage to 218W max from over 240W before but I'm still getting some thermal throttling.

There is even thermal paste coverage but I'll try checking the level as well as looking into the Adaptive Boost.

 

[Karadjgne]

AVX and linpack use in stress tests will guarantee raise temps well above normal ranges. Either one can push a cpu to @ 130% virtual at times depending on the stress test.

For temps, the Only viable test is Prime95 Small FFT (not smallest) with AVX tech totally disabled. That gives a clean 100% load, and since games do not use much, if any, AVX, it's a good baseline temp.

Aida64, occt, IBT etc can and do use linpack varients, along with AVX, so while good for stability, tend to suck for temps because the load is not stable, it fluctuates depending on the part of the test. Can be from 80ish% virtual load, to 130ish%.

 

[Frestyl]

Thanks guys. I've taken everything on board.
After all the changes I've made I was able to run a 10 minute cinebench R23. At the end of 10 minutes I was pulling 253W and never went past 94 on a single core. Previously I was thermal throttled after 30 seconds, even when only pulling 220W.

 

[Karadjgne]

Progress! 👍