VRAM usage on GPU devices
VRAM usage on GPU devices
I suspect the VRAM on my rog strix 3070 got damaged while I was installing K5 Pro. At first there were artifacts, so I lowered the GPU speed which resolved the problem until I raised the memory clock. I inspected the heatsink and confirmed it was making good contact; I added extra K5 Pro for backup, but when I tried to bump the memory clock again the system crashed. Now no memory is listed in gpu-z and there are still severe artifacts. I’m confused because everything appeared normal, maybe the VRAM overheated. Any advice would be helpful. iCloud Photos.zip
Hey there, artifacting often points to VRAM problems. I usually opt for thermal pads instead of paste, though there shouldn’t be any serious issues beyond a bit of mess. Right now, the best you can do is thoroughly clean all the paste and inspect the chips carefully for any visible damage. Remove every bit of residue, then gently brush with a soft toothbrush using distilled water. Replace the pads on the VRAM chips and VRM’s, and place the paste on the core die to check if the issue continues. That’s about it—I’d suggest getting ready for the worst case. Good luck!
shares content with a visible error about device manager problems, and launching basic apps triggers warnings about missing OpenGL version 2.0 support.
Oh wow, looks like a lot of trouble. Don’t lose faith, but there’s a slim possibility something in the paste could be the issue. Give it a thorough clean and check for improvements.
Consider using the older thermal pads if you want to keep the device running briefly without thermal pads. Running it without them for an extended period could be risky.
I'm not sure about this. Could someone share their experience? I tried putting my PC into BIOS for a brief time with a dry EK block on the CPU, but I didn’t attempt to boot a GPU without cooling. With GPUs, it's common they're mounted upside down or vertically, making it tricky to set up a temporary cooling fix during bench tests. You might want to look for some thermal paste on Amazon Prime and apply it to the core and pads. If you have a riser cable, you could place the cooler gently without tightening it, though even if the card is likely dead, a short cable won’t cause too much trouble.
I removed most of the debris and it’s still inactive—so it’s all gone. The card seemed a bit damaged from the beginning, likely used and crashed during stock clock time. I’ll probably wait for AMD to release a 4070 alternative since GPUs are really costly right now. Thanks for your assistance!