Problem with i9-13900KF - Core Thermal Throttling + Core Power Limit Exceeded + Package/Ring Thermal

[TaryfiX]

Hi all,

I just purchased my new PC and discovered that my processor has got some issues - I had so far BSOD 3 times.

I found a lot of articles about problems with 13th and 14th generation processors and decided to test mine with Intel tools.

I tested with:

Intel Processor Diagnostic Tool + HWInfo on the side to see more data

and next Intel Extreme Tuning Utility + HWInfo on the side.

Results are (in pictures)

Core Thermal Throttling

Core Power Limit Exceeded

Package/Ring Thermal Throttling

My setup:

13th Gen Intel(R) Core(TM) i9-13900KF 3.00 GHz

ASUS RG Z690 Extreme Gaming motherboard - Bios updated to latest version, Intel settings applied

32GB (2x16GB) Crucial Pro DDR5 6000MHz

Lexar NM710 Gen 5NVMe SSD 1TB M.2

PNY Nvidia RTX 4080 - 16GB DDR6

Corsair iCue Link H150i RGB 360mm AIO Water Cooler

Seasonic 850w Prime PX-850 Fully Modular 80 Plus Platinium Power supply

Windows 11 - Fresh install with all updates up to day

Is my Processor faulty and should be replaced?

I checked and is showing still on warranty till 2028

[PC1997]

As to why your PC is blue screening... there are many possible causes... since you said your PC is new. I'm going to assume you don't have much use time on the new PC? Did you build it yourself or is this a prebuilt?

Based on the screenshots you've shown, everything looks exactly as you'd expect from a typical 360 AIO and the 'performance power profile'. So you're aware, you are not using Intel's true default power profile:

You should be fine with the performance profile - PL2 & PL1 = 253 W versus Intel default, PL2 = 253 & PL1 = 125 W (Package Power Time Window = 56 secs) and both Intel and Performance profiles should be set to 307 amps.

You may be able to improve your temps by remounting your cooler and making sure you have thermal paste covering the entire IHS (butter spread method) and that your cooler is completely mounted - until the screws gently bottom out, using an X tightening pattern. I like to use two fingers with a screwdriver turned slowly to avoid applying too much torque.

The reason you want it to be so tight is because 13th and 14th gen processors do not sit perfectly flat due to extreme mounting pressure causing them to bow slightly. You will not see this with your eyes, but it will affect the way your cooler makes contact, depending on the surface shape of your cooler. The other alternative is to get a contact frame. Just make sure you're getting an authentic version if buying from Amazon etc. There are a lot of fakes out there that are made of plastic instead of aluminum. The plastic ones will not allow your cooler to seat perfectly flat and being made of plastic means it will flex - defeating the purpose.

Again, your temps are to be expected and you are running a synthetic workload, so real world programs will run cooler. It's a hot running processor and without extreme cooling (direct die) you are always going to be hitting the 100°C thermal throttle limit under any kind of substantial workload. Gaming workloads will see lower temps.

About the performance power profile; while you should not have any problems using it, theoretically, given known vulnerabilities of Raptor Lake processors, you will be running acceptable parameters as designed for the chip, except you will have no turbo boost time limit. Meaning your CPU will boost indefinitely limited by temperature / duration of the workload up to 253 Watts. Your CPU will not downclock to the base 125 W power limit under load. However, your processor should still down clock when completely idle.

The reason this MAY be a problem long-term is excessive current draw or acceptable power, but for an excessive duration may cause degradation of the clock tree circuits responsible for controlling frequencies of the different parts of the CPU. And as long as you have the latest microcode update 0x12B controlling for voltages, then the only other thing that causes them to degrade is excessive power draw.

It's not 100°C temperatures -  it's how much power are you drawing for that same 100c thermal throttle limit, e.g. 125 W and 100c, 253 W, 100c or 300+ (over boost) and 100c. And of course, if you set max thermal throttle limits over 100c that will cause a problem by itself. The point is: your cooler limits how much power your CPU can pull before hitting its thermal throttle limit. So the more amps you pull and the longer the duration, the greater the chance of degradation.

If in doubt, use Intel's actual default power profile, paired with the latest BIOS update.

[Greebo14]

I`m the PC builder thought. 

The PC went through 12 hour test with Prime 95 and it never went up 86C for a mere a second. 

I handled it with the customer and assured them that everything was fine. It was updated with the latest bios, etc. 

As you mentioned - the tests are synthetic and there is no such load in the real world. 

The BSOD is page_fault_in_non_fault_area - ntfs.sys which is often Windows-related. Overheating will cause an immediate shutdown. 

[TaryfiX]

Hi,

Thanks for reply and advise about contact frame.

Please see my Bios setting in the pictures attached - Intel Default settings are On and PL1 and PL2 are set on 253 Watt

CPU core/ Catch Limit Max - 400A

Time window - 56 sec

Bios version - American Megatrends Inc. 4101, 03/12/2024 - Latest one

Please see also 2 tests results while playing on Fortnite from HWInfo

When I only use the browser and HWInfo right now, my temperatures are between 44-72 degrees

Just did tests on Prime95 and it's red  (picture attached)

[PC1997]

1. You are actually using the extreme profile and not the performance based on screenshots of BIOS. Reset to defaults and with the latest bios you already have should load the correct parameters. Make sure your current limit is set to 307A, not 400A and retest for stability.

2. If you get the same BSOD error, disable XMP - retest.

3. If necessary run MemTest86 in BIOS with 1 DIMM (if pass) 2 DIMMs. Also check the QVL list to see if your memory is compatibility with the board.

4. Check Windows Event Viewer for WHEA errors, specifically 19 error events.

5. Check for instability problems when compiling shaders with Unreal Engine 4/5 games. Black Myth: Wukong Benchmark Tool is free if you don't have any Unreal engine games to test. If you do have games to test, updating your graphics card driver will force shaders to recompile when launching games.

You can save stability testing time by running Cinebench R15 / R15 Extreme. You can pass other stress tests and fail with this one.

BenchMate is an excellent source of stability testing programs in one convenient software suite and is free.