[SOLVED] Game crashing randomly when under load

[TheArchonPrime]

I recently got this computer about 3 days ago, and i usually played D2 on it, when it came to everything that isnt a high load, it ran it perfectly fine. However in areas that max out the graphics card for and extended time, like 10-20 minutes, the game will just randomly crash giving the error "graphics runtime detected a crash or loss of device"
i did look at the temperatures and they went no higher than 65c at any point.
When I tested with GTA V, the whole computer straight up blue screened after about 40 minutes

what I found particularly odd is that when i ran it under Kombustor for an hour, it was perfectly fine with no issues

Specs:
Intel i5-12400f
RTX 3070 Ti
4x 8gb DDR4
ASUS PRIME Z690
750w gold PSU

 

[rgd1101]

brand and model of the psu?
is this a ibuypower or cyberpower?

 

[Dirtmechanic]

Is the bios setup for the larger chunks I think default is smaller it makes a bottleneck just curious

 

[TheArchonPrime]

brand and model of the psu?
is this a ibuypower or cyberpower?

ibuypower

CORSAIR RM750X - 80 PLUS Gold, Fully Modular

gonna try and set a power cap on my GPU and see if that solves it, if it does i'll upgrade my PSU, ill edit if it works or not
result: at a 75% power cap it still went through issues, which makes me think it might be a faulty GPU, ill test again with 50% tomorrow

 

[smtrejo]

I've had similar issues whith a Zotac RTX 2060 6GB on the last 3 weeks, it crashed on AAA games or heavy load (specially on DirectX 12) but it passed all benchmarks and stress tests. I tried every tweak/workaround I could find and nothing worked. But finally I found a step-by-step guide on YouTube to modify the gpu's voltage/frequency curve (undervolt) with MSI Afterburner:

View: https://youtu.be/--YvuxE0xl4


In my case I had to undervolt to 900mV (0.9V) and 1725Mhz (the video locks maximum voltage to 1v, but you'll have to trial and error as I did) and the games aren't crashing anymore. I played Cyberpunk for 3 hours whitout crash and some sessions in Fortnite (both games in Ultra/Epic and Ray-Tracing enabled). You should give it a try.

 

[TheArchonPrime]

I've had similar issues whith a Zotac RTX 2060 6GB on the last 3 weeks, it crashed on AAA games or heavy load (specially on DirectX 12) but it passed all benchmarks and stress tests. I tried every tweak/workaround I could find and nothing worked. But finally I found a step-by-step guide on YouTube to modify the gpu's voltage/frequency curve (undervolt) with MSI Afterburner:

View: https://youtu.be/--YvuxE0xl4


In my case I had to undervolt to 900mV (0.9V) and 1725Mhz (the video locks maximum voltage to 1v, but you'll have to trial and error as I did) and the games aren't crashing anymore. I played Cyberpunk for 3 hours whitout crash and some sessions in Fortnite (both games in Ultra/Epic and Ray-Tracing enabled). You should give it a try.

ill try this, thank you so much

edit: it doesnt appear to have worked, i tried capping my GPU at 0.9v 0.95v, 1v, 1. 025v, 1.05v and 1.1v, none really appeared to work, when i went below 0.95v the whole pc would straight up freeze

 

[smtrejo]

ill try this, thank you so much

edit: it doesnt appear to have worked, i tried capping my GPU at 0.9v 0.95v, 1v, 1.05v and 1.1v, none really appeared to work, when i went below 0.95v the whole pc would straight up freeze

Someone told me it could also be the gpu's BIOS, and the way to fix it is to "reprogram" it, so I suppose one has to "reflash" the firmware

 

[TheArchonPrime]

Someone told me it could also be the gpu's BIOS, and the way to fix it is to "reprogram" it, so I suppose one has to "reflash" the firmware

from what i can tell nvidia nvflash doesnt support the rtx 3070 ti yet

 

[TheArchonPrime]

FOUND THE SOLUTION

So it turns out, it was not only a faulty GPU, but faulty RAM as well
I RMA'ed the GPU which completely eliminated the blue screen crashes, and i ran prime95 and found that the memory was weird too, so I underclocked it to 3400 mhz from 3600 and everything seems stable now