My AMD 7900 XTX crashes all the time running LLM's and Stable Diffusion. With ROCm installed correctly. Paired w/ Ryzen 7 and 64g Ram. On Arch Linux. So why does AMD's top of the line GFX card need almost 5 hard reboots every day? Sometimes more. It feels like my RX 6800 & RX 570 ran better.
My build:
Chassis: PHANTEKS G500A - BLACK
GPU: AMD RADEON RX 7900 XTX
Processor: AMD Ryzen 7 5800X3D
Motherboard: ASUS TUF Gaming B550-PLUS WiFi 2
Memory: 64GB META DDR4 2400MHz (4x16GB)
Cooling: DEEPCOOL AK620 - ZERO DARK
Power Supply: META 1000W
Fans: Bequite Silent Wings4 140MM Fans x 6
Operating System: Arch Linux (Not a n00b)
So I clearly have a decent setup. So why? Why does it seem this gfx card struggles so much? What is with the constant crashing or freezing up needing a hard reboot? I have no clue on where to begin fixing this expensive piece of hardware.
I choose AMD all the time when I have the choice. But, I'm starting to regret it. Is there a guide to the best settings? Maybe environment variables I'm unaware of? Anything?
I'm willing to put in the work and compile whatever from scratch without help. Just point me to some documentation and source code. I'll get it done.
Anyone?
Can anyone help me, help myself?
Thank you for taking time to read this. You didn't have to. We're all on this Earth for an unknown amount of time. And I appreciate you spending some of that time reading my post and possibly replying with a solution for me. Thank you.