Error: Couldn't switch to requested monitor resolution about dxvk

[K0bin]

Thanks for reporting back.

[doitsujin]

err: D3D11SwapChain: Failed to recreate swap chain: -13
err: DXGI: CreateSwapChainForHwnd: Failed to create swap chain, hr -2147467259

Looks like some problem with PRIME maybe? For some reason we're unable to create a Vulkan swap chain on the game's window, but the reason for that can be pretty much anything from messed up driver setup (shouldn't really be a problem since Turing should just work out of the box), your wine/proton build being weird, Heroic setting weird environment variables that fuck things up, or the specific game doing something weird. Extremely hard to remote-debug these things.

Maybe try something like DXVK_FILTER_DEVICE_NAME=Intel to force it to run on the integrated GPU and see if that works for now.

[BaldousHuxley]

I don't really know about the drivers, installed smoothly from rpmfusion and signed for safe boot. I already reinstalled them once to be sure a few days ago, but the issue persisted. Setting DXVK_FILTER_DEVICE_NAME=Intel does run the game, but of course there are performance hits on more demanding titles (for the record, almost every game I tried firing up from both Steam and Heroic throws this error).
Since remote-debugging is not viable, how would I go about investigating it? I'm pretty new to Linux so I'm just asking about some good debugging practices. Also, in the event I manage to launch the game, would the log be of any use? Thank you for your patience. I'm kinda going insane with this

[BaldousHuxley]

Hello, with the help of some (way) more knowledgeable people (than me) I found a solution for anyone having the same issue. Basically the root of it was that during boot, one of the Nvidia kernel modules failed to load. In my case, nvidia-modprobe failed to initialize, and as a result, some key system files were missing at startup; for me it was the /dev/nvidia-modeset device file. The curious thing is that vulkaninfo fixed the issue because it directly attempts to call nvidia-modprobe after the boot initialization (as explained well here and also here). This means that running vulkaninfo deals the same exact result as running nvidia-modprobe -m (to load the modeset module and create the missing file). Now, the easiest and preferred way to automatically run nvidia-modprobe when the Nvidia module is loaded during boot, is by setting a custom udev rule: the structure of which is almost identical to this one made for Arch. There are some personal differences in the KERNEL and DRIVER arguments based on how each is called in the host machine, so let's go step by step:

0. You may want to check if you have nvidia-drm.modeset=1 in the boot parameters. If not, it is preferred to add it (there are plenty of resources online on how to add a boot parameter in GRUB, I don't wanna make expansive explanations. Also note that depending on the distro of choice, the spelling could need a _ instead of - so check for your distro);
1. Create a new file in /etc/udev/rules.d called nvidia-modprobe.rules;
2. Copy the contents of the Arch rule in the newly created file;
3. Run udevadm info -a -n /dev/dri/cardX, where X is the number of your dGPU. Should be either 0 or 1 (you should be able to infer which card the command targeted by looking at the output anyways);
4. Find the right KERNEL and DRIVER entries listed by the command and replace the default ones in the nvidia-modprobe.rules file with those (in my case they were KERNEL=="card1" and DRIVER=="");
5. Add the right flags to the line RUN+="/usr/bin/nvidia-modprobe -c 0 -u". I, for example, had to add -m to rebuild the right missing module (modeset);
6. Save the file, reboot and you should be good!