14900K Instability (un)SOLVED

[CatTNT]

I got a 14900K in December of 2023 with a really good deal from microcenter, $700 for a 14900K, 32GB 6000Mhz DDR5, and an MSI Z790 Tomahawk MAG motherboard. 
Ever since then, I've had TONS of issues with stability, and literally nothing I did could possibly solve it. It didn't matter if I could run an Aida64 (small ffts, and lots of different variations) stress test, play games for hours, or do anything else, my whole computer would eventually crash after 1-12 hours. I have tried EVERYTHING:

* Reset Windows (While keeping files)
* Disable all startup programs
* Uninstall and reinstall Vanguard (Riot games anti cheat)
*Remove and undo all overclocks for GPU, RAM and CPU and set all to defaults. 

*Change NVIDIA driver version (I have used 4 separate versions, 1 in the 400's range the rest in the 500's range)
* Run system file checker multiple times
* Run DISM multiple times
* Upgrade AND downgrade BIOS (i tried using the new Intel Defaults that came with the updated BIOS, still doesn't work)
* Remove 1 out of 2 sticks of RAM, and swap which stick is in use

*Get a new 4TB system drive
* Completely reinstall windows on that new drive and reinstall the bare minimum of programs I need to use
*Enable windows driver verifier
*Ran windows memory verifier (failed after going about 10% through, computer totally froze up for hours)


This is about 2/3rds of the things I tried to fix the problems and none of them fixed it. My errors included a pool of around 5-7 different types of BSOD's,  at my worst having 10 BSOD's in 3 days. They were all said to be memory, windows, or program specific problems, but none of them ever said anything about the CPU. My theory was that it was eitehr my RAM, PSU, or Mobo. The specific bluescreens were: KERNEL_SECURITY_CHECK_FAILURE, DRIVER_IRQL_NOT_LESS_OR_EQUAL, YSTEM_SERVICE_EXCEPTION, ATTEMPTED_EXECUTE_OF_NON_EXECUTE_MEMORY, DRIVER_VERIFIER_DETECTED_VIOLATION, KMODE_EXCEPTION_NOT_HANDLED (vgk.sys, and 1-2 more I didn't take a picture of. In addition to that, my PC couldn't start at least once with the error code of 0xc0000001. I  also couldn't reinstall my drivers for a time because i got a 7zip CRC error every single time I tried to open the NVIDIA driver installer. Basically, the whole system was a shitshow and nothing was reliable or stable and there was NOTHING I could do about it. Until I got a hunch after watching a video about the unreliability of 13900K/14900K CPUs. All I did was reduce my Performance (P) core frequency by -200 Mhz  in the BIOS and there's been zero issues since.   (P Core turbo ratio offset -2)  I've got all my programs back, played games, mined, ran a server, overclocked my 4080, (NOT THE CPU), and everything has been perfect. Sure, my max speed is 5.8Ghz instead of 6Ghz, but at least there are NO issues now. This has been stable for at least 2 weeks by now. Before, my computer couldn't even stay on while doing nothing for 24 hours. 

CPU Settings
All default except Vcore: 1.42 volts, and P core turbo ratio offset -> -2

HWINFO shows my max Vcore is about 1.422 volts, max CPU speed is 5785 MHz

EDIT: kept the speed nerfed I talked about above, but reduced the Vcore to 1.41, still works great after a couple of days, no crashed no errors
Also, my CPU package spikes to 107C according to HWINFO, but no core temp has gone above 100C. This is probably a big issues, but it looks like 100+ temps are something that occurs for just a second or two at max. Usually with highly threaded workloads I'm getting 80-95C for a couple hours. 

Edit 2: After reducing the P core offset to -3 or -4 and voltage to 1.39V then going on vacation, It has bluescreened at least twice. I'm not sure if its actually stable anymore or if that was just a fluke. I'll post updates after flashing the BIOS with a microcode update and using it for a couple weeks. 

Edit 3: After underclocking by -2, keeping all E cores enabled, and removing XMP, the CPU seems to be stable. It looks like RAM speed is another area of instability. At 4800MHZ, the CPU's SA voltage is 0.81. At 5200MHz, the SA voltage is 1.151. I suspect that at high enough SA voltage caused by using high enough RAM speed will cause crashes, along with degraded CPU cores being another issue.
The system is currently appears to be stable with 1.39 Vcore, -2 P core offset, XMP disabled, and running the 2nd LOWEST version (7D91vH9)  of my motherboards BIOS that supports 14th gen. I'll try slowly increasing my RAM speed to see what's the maximum that's stable. 

[Fasauceome]

Gamers nexus and Wendell from Level1techs just posted that exposé about Intel's problems at scale with 13th and 14th gen. Glad you are not having issues anymore. 

[eece_ret]

Im curious if we are seeing something similar to https://www.theregister.com/2017/02/06/cisco_intel_decline_to_link_product_warning_to_faulty_chip/   

[Ebony Falcon]

All you have to do is get rid of the single core boost and lock the cores, if you still crash after that it’s probly slightly degraded, drop clocks and extra 100mhz,

there mostly crashing becuase of the voltage it shoves for the single core boost then voltages dropping back for multicore