Question Pc crash supposedly because of GPU heat

Aug 16, 2023
3
0
10
Hi everyone, I recently started generating images using AI and more precisely Stable Diffusion. It's installed locally and ran quite smoothly until now. I keep having crash (i.e. black screen, GPU fans at max speed, buggy sound if there was some music playing at the moment) when I generate even small images.

Since the situation occures when GPU (MSI RX580 8GB OC) is at 100% use and near 86°C I supposed it was because of cooling. I bought some thermal paste because the GPU wasn't brand new ande I thought it may be dried and the vendor sold me a liquid metal one with 3.07W/m-K conductivity. I was surprised because usually gamer pastes are near 12W/m-K.

Nothing changed so I want to ask you some advices about how to cool my GPU more efficiently without using watercooling (it's expensive and I'm not at ease with the maintenance of such system).

Currently I have a 140mm BeQuiet fan on the front, and a smaller one (maybe 90mm, I'm not sure) on the back. My motherboard cannot hande more fans because it only has one slot for it (I bought a Y cord to connect both on the fans to this slot).
 
Welcome to the forums, newcomer!

because the GPU wasn't brand new
So how old is the GPU since it's been in your hands? How long has the GPU been in service prior to your purchase? If the card was used for mining, and your PSU is not at fault, then you're only option is to return the card or live with it as is.

As for your system, please list the specs to your build like so:
CPU:
CPU cooler:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:
Monitor:
include the age of the PSU apart from it's make and model.
 
Welcome to the forums, newcomer!

because the GPU wasn't brand new
So how old is the GPU since it's been in your hands? How long has the GPU been in service prior to your purchase? If the card was used for mining, and your PSU is not at fault, then you're only option is to return the card or live with it as is.

As for your system, please list the specs to your build like so:
CPU:
CPU cooler:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:
Monitor:
include the age of the PSU apart from it's make and model.
Thanks for the quick reply. Here are my specs :

CPU: Ryzen 3 3200G
CPU cooler: Cooler Master, maybe a i30 or i50 (sold with the case)
Motherboard: Gigabyte B450m DS3H v1
Ram: 2x 8GB Corsair DDR4
SSD/HDD: 1x Kingston SSD 250 Go
GPU: MSI RX580 Armor OC 8GB
PSU: Textorm TX500+ 500W 80+ Bronze
Chassis: Cooler Master MB511
OS: Windows 10 (up to date)
Monitor: Dell 24" ST2420L

Maybe an important info : due to bad VRAM management by AMD "gamer" driver, I installed the studio pro one. It should be up to date too.

Apart from the monitor and GPU that were already used and sold to me, every component was brand new and bought in 2020. So PSU is 3 y.o.

I don't exactly know when the GPU wasn't bought but I'm sure it wasn't used for mining. The guy who bought it barely used it because of a slot issue between the motherboard and the GPU when he upgraded his pc. It sold it to me in March 2021.

Voila !

Edit : How may I be sure the PSU isn't faulty ?
 
Unless you have some pretty high-end equipment, the only way you can test the PSU is by swapping in a known, good one. Hopefully one of better quality than the cheap, group-regulated unit you have in there.
I plan to move to a very better config, including a RTX 4090 24GB, for this I would need a 1000/1200 W PSU. Would it be any issue to upgrade the PSU now, will it consume more power or stick to the real-time consumption ?
 
I plan to move to a very better config, including a RTX 4090 24GB, for this I would need a 1000/1200 W PSU. Would it be any issue to upgrade the PSU now, will it consume more power or stick to the real-time consumption ?

PSUs only use the power required; a 1200W PSU doesn't use more power to power a GPU than a 500W PSU does to power the same the GPU.

If you're getting a quality PSU anyway, it won't hurt at all to get it now.