Question PC crashing just during 3D games played

[ramireus]

Hey guys,
I would really appreaciate help before sending my PC to repair shop, as I feel I am missing something. My PC was working hapily for 4 year and recently started crashing 0 - 30 minutes after playing 3D games (Fallout 4, Everspace, XCOM2, Stellaris) at random points of the game (sometimes at startup (more common for XCOM 2), mostly later). Weird is, that 2D games are unaffected (played Rimworld for cca 5 hours straight without any problem), as well as any other work.

Here are my Specs:

• MB: ASRock Z170 Pro4
• CPU: i5 6600K
• GPU: GIGABYTE GTX 980 Ti G1 GAMING (driver 431.18)
• RAM: Crucial Ballistix Sport LT 8GB (2x4GB) (Yup, I know, I plan to increase it)
• PSU: EVGA SuperNOVA 750 G2

What I tried so far:

• Clean W10 reinstal
• DDU driver removal and reinstalation
• all drivers up-to date
• older GPU driver
• claning RAM memories from dust
• MemTest, GPUMemtest, furmark burn in (cca 15 minutes, GPU is periodically throtling speed, I red it is normal for this test), 3D Mark stress test, Prime95 stress test, Mark10 extended test-all without error, temperatures kept low
• Playing from different HDD and SSD

Clues:

• When crash happens, the computer freezes with screen turing uniformly one random color, reacts only to reset
• Sound sometimes run correctly for few seconds, then crashes as well

I will be glad for any ideas, where can be problem,
thank you

 

[PC Tailor]

GPU is periodically throtling speed, I red it is normal for this test

Not necessarily. What GPU temps are you hitting?
Also what CPU temps were you hitting (in individual cores, not the whole package).

if you've clean installed and the issuye has persisted, it is more likely to be hardware.
Do you also have latest BIOS installed?
How long did you run memtest for?

 

[ramireus]

GPU temps maxes out at 66°C under furmark-cyclic throtling at this temperature
CPU temps on cores shown by speedfan temperatures do not cross 70°C for all cores (each watched separately) during 3 hour Prime95 testing
Stressing GPU and CPU at the same time show no issue
memtest was running for 3 full passes, ended during pass 4, all tests, cca 1,5 hour
BIOS was updated last time one and half year ago, no hardware change since then, but I guess it is worth a try.

I am afraid you are right it is probably hardware related :/ I just keep hoping it is RAM and not something serious. Basically I can rule out PSU only

One more thing I noticed is that GPUmemory is throtling during furmark drastically (from 3000+ MHz to cca 800MHz)-this can be possible culprit?

-Another note-3D games crash even on minimal settings

EDIT:
clarifications

 

[PC Tailor]

Basically I can rule out PSU only

Why can you rule this out?
Faulty PSUs can quite commonly cause this issue.

It may also be worth seeing if the freezing still occurs if you remove the GPU and run on integrated on minimal settings.

 

[ramireus]

Why can you rule this out?
Faulty PSUs can quite commonly cause this issue.

well, I think that such problem would show itself when I tested simultaneously full CPU and GPU stress (Prime95+furmark or 3Dmark). I cannot imagine how low setting game can stress out PSU more-but I might be wrong.

Will try to remove GPU probably tomorrow.

Thank you for tips

 

[ramireus]

It may also be worth seeing if the freezing still occurs if you remove the GPU and run on integrated on minimal settings.

Had slight emergency and had to postpone further testing. With removed GPU the crashes stopped, therefore I am afraid that the most expensive part went to hell. Will try to play with it a little in Afterburner (eg. speeding fans up in case it is overheating od some part)

Anyway, made full cleanup, new thermal paste for CPU, off course it did not helped.

 

[ramireus]

well, here is result of test, also maybe one thing I have not stressed enough is visible from this picture: Periodic throtling of GPU, with no extreme temperature

https://imgur.com/4ZnSQGL

View: https://imgur.com/4ZnSQGL

 

[PC Tailor]

I cannot imagine how low setting game can stress out PSU more-but I might be wrong.

It's not necessarily about stress on the PSU, the PSU regardless has to maintain regular and stable power to other components, it only needs to have a slight fluctuation of voltage or ripple and it can cause crashes at high or low loads.

Did you mention that the crashes completely stopped with the GPU removed? It may be worth continuing to test this, if it can pass a stress test, it's just indicating it worked at that time.