PC shuts down under high GPU load

[rdmchr]

Hello everyone,

bit of a back story: I recently upgraded my GPU to a second-hand RTX 3090 and the first day with that card was great. But on the second day, I noticed my system was getting relatively slow so I used HW Info to find out that my CPU was running hot. 
The problem seemed to be an issue with my Lian Li AIO (see this blog post), so I requested a new AIO which I have yet to receive. 
In the meantime, I installed the stock cooler that came with my CPU. This configuration seemed fine for the evening, but when I tried playing some Doom Eternal tonight my system suddenly shut down.
I'm not sure what really happened, but my monitors went black, even though the little lamp in the power-up switch on my case remained lit.
After about 2 minutes the system restarted and I tried playing Doom again, this time keeping HW Info on my second monitor since I felt like my case was getting a bit toasty.
I also put a fan in front of my case and removed the front panel to help with ventilation. But after just a few minutes of gameplay, my monitors turned off and then back on again, this time with some weird artefacts (see attached picture).
After one minute my monitors turned black again and the system began to reboot. I quickly checked HW Info before that and everything seemed "fine"; temperates were in the 60°C range.
So I installed Furmark and ran a GPU stress test and at 56°C my system turned off. (not sure if that helps, but I got about 200 FPS in 1440p preset)
I'm not quite sure what I should do here. The seller seemed trustworthy and I picked the GPU up at their place.

 

Some additional information about my setup:
- my current cooling situation is bad; I'm using the Lian Li O11 Mini case with just three fans at the top as exhaust fans. My AIO was mounted on the side of the case, also as exhaust. I'm aware that this is bad and I already ordered new fans that I want to add as intake at the bottom. Furthermore are the fans I currently use Noctua fans which according to their site are meant to be used with radiators.
- on the day I noticed the malfunctioning AIO, I'm pretty sure I caused some thermal damage since the system was already running quite slow (probably due to overheating) and I accidentally ran a Windows update. (the update did seem to finish without a problem though)
- my case blocks the HDMI port of the GPU, so I mounted it at a slight angle at the beginning. I have since fixed this and ordered a bolster for additional support, but that won't arrive until the end of the month. There seems to be no damage to the PCI-E slot

 

System Specs:
Motherboard: MSI MAG B550 Tomahawk (BIOS Version: E7C91AMS.A50, Build Date: 2021-01-15; I should probably update this? But Bios updates scare me )
CPU: AMD 5 5600X (I ran this auto-overclocked using the Ryzen Master app until the AIO died)
GPU: RTX 3090 iChill by Inno3D (Driver version: 537.34, updated via Geforce Experience on the 12th of this month)
PSU: Corsair SF750 (750 Watt)
Storage:
    - 1 Samsung 980 Pro 1TB
    - 1 Samsung 870 Evo 2TB
    - 1 Crucial MX500 500GB
RAM: G.Skill Ripjaws V, 32GB DDR4-3600 (dual 16GB)
OS: Windows 10 Pro Version 22H2 (OS Build 19045.3448) 64Bit

 

Software:
I'm not 100% sure, but this should be all the software that ran during the crashes:
Doom Eternal Crash:
    - Spotify
    - Docker Desktop
    - Wallpaper Engine
    - Discord
    - Steam
    - Epic Games
    - f.lux (https://justgetflux.com/)
    - Microsoft Powertoys
    - Ear Trumpet
    - Logitech G Hub
    - MSI Dragon Center
Furmark:
    - Furmark
    - Ear Trumpet (forgot to stop this one)

 

Hope I didn't ramble too much, but I think the path my system took in the last few days might be important here. Thanks for any help/tips.

 

TL;DR: System crashes when GPU reaches ~60°C, sometimes with pink artefacts.

[Middcore]

Artifacts like this are virtually always a symptom of a GPU hardware failure, in my observation. Often related to the VRAM. This looks really similar to the VRAM issue on 20-series cards that people called the "space invaders."

 

You said the card was bought secondhand. Do you have any recourse to return it? 

[m1lk3y]

23 minutes ago, Middcore said:

Artifacts like this are virtually always a symptom of a GPU hardware failure, in my observation. Often related to the VRAM. This looks really similar to the VRAM issue on 20-series cards that people called the "space invaders."

 

You said the card was bought secondhand. Do you have any recourse to return it? 

I've seen enough videos on YouTube from northwestrepair to know that I can comfortably agree that it looks like artifacting probably caused by a bad memory chip. If you're outside warranty, I'd check his channel out, I don't think I've seen a video of his where he wasn't able to fix a card, dude is a wizard with repair.

[rdmchr]

On 9/14/2023 at 11:29 PM, m1lk3y said:

I've seen enough videos on YouTube from northwestrepair to know that I can comfortably agree that it looks like artifacting probably caused by a bad memory chip. If you're outside warranty, I'd check his channel out, I don't think I've seen a video of his where he wasn't able to fix a card, dude is a wizard with repair.

Alright. Thanks you two for the swift response. I have since send the card to a local repair shop since I don't have the expertise or equipment to do the repair myself. Let's hope it's just the VRAM that's fried and that they can repair that.