I9-13900k, stable in daily use, games, etc. Crashes under benchmarks and stress tests

[Glement]

Have been daily driving my system for 1 year, 
Only experienced crashes in 2 games (soft crashes, not bsod) - starfield on release, and crossout.
Also have a 2 games combo that usually bsods my system after 5 hours of them both running ( not always, sometimes they can run for 10 or more without bsod) - they are Escape From Tarkov and Black Desert Online, if i play them by themselves, nothing wrong. 

Those crashes in games i attribute to games being kinda bad coded. 

Lately i wanted to run some benchmarks and stuff to see how is my cpu doing. And i found out that my cpu does not pass any cpu intensive benchmark,
cinebench23 - either soft crash or bsod.
intel avx2 test - soft or bsod
prime95 - soft crash 

Looking on this forum i found out that some people had issues with 13900k, and the issue was that they had SVID on Auto, and the recommended solution was to put it into Intel Fail Safe

After having SVID on Intel Fail Safe -> i pass all the benchmarks, but i run hotter, and in prime 95 my P cores are at around 4.5 GHz, cinebench around 4.8 GHz and thermal throttling.

My question is - if under normal load i do not crash or experience any bugs - is it still safer to have SVID on Intel Fail Safe? 

Tests were made on Z790 Maximus Hero with default BIOS with following settings
MCE - disabled, enforce all limits
ABT - off
XMP - profile II or auto (stock clocks)
SVID - auto or intel fail safe

Memory - GSkill Trident Z5 2 sticks of 16 gb, 7200mhz cl34

NH-D15 from Noctua for cooling, with 2 fans
Thermal Grizzly Kryonaut Thermal paste

 
my VIds with Intel Fail Safe
And temperatures running Black Desert Online 

[Glement]

delete this reply

[vmovups]

Can you send crash dumps from %localappdata%\CrashDumps folder?

It could help both the Intel staff and me to understand if everyone's CPU is crashing in the same manner.

Has the problem also occured from day one or was it gradual? If it came gradually the Intel Fail Safe option may worsen the problem as it is increasing the voltage to compensate the CPU degradation but the additional voltage is degrading it quicker... (in the screenshot you provided the vcore maximum was an uncomfortable 1.598V)

Depending on the game and the exact code it's normal that some never crash while others consistently expose the CPU's instability.

Do not hesitate to RMA the CPU or even try getting an exchange/refund, the problem is too common to be ignored.

[Glement]

0. I haven't run those test when i built the pc, the only thing i ran when i built it was furmark to check temperatures, and it was stable.
1. Can you specify which CrashDumps you want?
2. As i said, after building my pc, i didn't do anything except enable xmp II and disable the mce and abt, games were perfect, nothing wrong, the only issue i have is benchmarks as i said. If i put SVID on AUTO and start playing like cyberpunk on max settings or some other AAA title, i do not see any crushes NOR bsods, the only game causing me to bsod was combination of tarkov+black desert online, and maybe starfield, but starfield might be raw on release so i am not sure about starfield.
And considering that stresstests that use avx alot cause it to crash on SVID Auto, i guess the problem could lie in AVX territory? 

If intel offers RMA, would probably agree to it. 

[vmovups]

it's possible that having ABT disabled is what is preventing the games from crashing, the CPU is supposed to be stable with ABT and MCE if they're the motherboard defaults but crashing/BSODing with both disabled is certainly odd...

Try to see with the shop/distributor you bought it at if they can do anything, they may want to replace the CPU after reproducing the issue at default settings.

[Glement]

I told you, crash happens ONLY in benchmarks when SVID is on auto
i haven't had crashes in normal use apart from what i described with SVID auto

[Glement]

forgot to mention that i played fine first month or two with mce and abt on
because i didn't even know about them
i disabled them later because they caused spikes to 95-99 c
without giving any difference in performance 

[Glement]

Here are the files, i excluded 3 files which are related to me updating drivers, so crash during driver update is expected. 
Crashes are from today, i do not know why i do not have crashes from yesterday when i did all the testing. 

[vmovups]

I cannot see the code that crashed as i don't have the crashing program and it appears to be a UWP app but it always crashed in the same code which could be a sign of both a software bug or the hardware problem.

The best way to tell which one it is would be to try to see of the creation of crash dumps from this app stops when the windows power plan is set to energy saving and starts again when normal or high performance power plan is used. 

If the app didn't crash before and started crashing today despite nothing having changed, it is not a good sign.

[Glement]

But i do not see any apps crashing 
what exactly are you saying / want me to do? 

[Glement]

here are some more dumps when running svid auto 

[vmovups]

The Tarkov dumps are interesting in that in both the CPU did not appear to be executing valid code(but i don't have the game installed so difficult to 100% confirm), which appears to match one of the crashes i had on my replacement 14900K.

From this, the chrome crashes when watching videos and the fact it stops when SVID is changed i'd say it's safe to say your CPU is affected.

The SVID values are normal on Auto, but if the CPU degraded on Auto it will probably degrade on Intel Fail Safe too...

[Glement]

So your replacement cpu has the same exact problem as the one that was replaced? 
That is unlucky 

[Glement]

Just had this in chrome when watching a youtube video
I am on SVID auto currently 

[Glement]

My VID per core on SVID - auto
and watching videos/discord with high performance power option 

[SHAB]

There have been issues with the 13th gen and 14th gen cpus.

Try setting the P-Core Ratios to 52-54. I think the factory overclock settings is making the system be unstable so you experience random crashing and unstable applications. The P-Core ratios are like what you want as the new maximum to be 52 = 5.2ghz 

I had this problem and solved it by de-clocking the cpus to a new max instead of factory max settings.

This could be due to drivers from oems like from motherboards bios and windows update itself. When the 13900k was released there was no problems, so it could be due to the software not playing right with the highly tuned factory overclocked settings.

[Glement]

it would not solve it, i think, when i enable intel fail safe - it downclocks to 4.8 when performing cinebench test due to thermals
if i go with auto - can't say, since i crash benchmarks, will try 5.2, but still, it can be the degradation overtime as most people suggested here.  since i haven't had problems in daily use, and i can't confirm that i had problems with benchmarks when i built the pc because i haven't run those benchmarks when i built it. 

[SHAB]

don't enable intel fail safe, turn off intel turbo boost and see if it gives no problems in benchmark app then if it doesn't then put just ur p-ratio to 52-54x and then test again.

[Glement]

it is still only hiding the problem
and i guess if it is progressing from the day i started my pc for the first time, soon enough i might be running with turbo completely off xDD 
will try it today 

[Glement]

i tried with svid auto, no turbo - cinebench 23 and avx2 from xtu passed
tried svid auto, 52 sync all p cores - crash
tried svid auto, 50 sync all p cores - works
Will wait on what intel advises. Because using svid on intel fail safe is REALY bad in terms of what voltage i see and what temperatures i see. 

[SHAB]

Nice, always look for system stability first before anything. I'm running a KS chip so its a bit more binned for more leeway with the P-cores. If you are using the regular chip its a lottery to see which portion its stable at. This way you don't lose most of the cpu's power. Worst case Scenario is you RMA the chip but be warned you might have no computer for like a month unless you buy a 13100 cpu as a basic backup chip.