When my GPU crashed

My primary workstation has an ASUS TUF X299 MARK 2 Motherboard which is almost ten years old. This motherboard does not support a processor that has an on-board GPU. As a result, I’ve installed several GPUs over the years, changing or upgrading as I bought new displays that supported higher resolutions.

The GPU I’ve been using for the past several years is a Radeon Pro WX3100. First released in 2017, it’s been a powerful yet relatively inexpensive GPU, and has met all my needs for the last several years. However, my workstation runs 24×7, I usually run many applications on multiple desktops, including multiple virtual machines for experiments I use in my books and articles. So it gets a good workout.

The problem

The problem revealed itself quite dramatically when my screen went black when trying to wake up from sleep and, after a couple long minutes, reverted to 640×480 resolution, and only displayed a hugely magnified section of my desktop. The desktop was unusable as I couldn’t access any desktop tool to reboot or power off.

I tried Ctrl-Alt-F3 to go to another virtual console where I tried to change to the multiuser.target but that failed. A reboot worked — until the next time this failure occurred. At that point nothing worked, so I logged in from another host on my network and rebooted from there.

After that, reboots never finished. I was left with a black screen, and I couldn’t login remotely because the network hadn’t yet started.

The solution

At first I thought the display might be the problem, but it worked fine through the EUFI BIOS boot and the very early stages of Linux boot through GRUB and a bit of the kernel startup until it tried to reset the display to a higher resolution.

The next obvious culprit was the GPU. Fortunately I keep a supply of spare parts, so I had another identical GPU in a drawer. After installing that everything works fine.

Summary

GPU problems of any kind are quite rare. This wasn’t the first time I’d had a GPU problem, but it’s been years since I have. It started and progressed through increasingly severe symptoms over a time span of a couple days, to the point where the system was unusable.

The fix was simple: replace the GPU.