| 5 Feb 2024 |
 @atemu12:matrix.org | Perhaps WINE might be doing some device selection thing where they try to not use iGPUs if a dGPU is present | 05:54:54 |
| @atemu12:matrix.org | I've heard of some cases where the opposite was the case | 05:55:13 |
| @atemu12:matrix.org | So that may be a workaround for that | 05:55:24 |
 ersei | seems like vkcube doesn't want to run either without that variable set too | 05:57:38 |
| ersei | actually it launches, but with a 20 second delay | 05:58:22 |
| ersei | seems like vkcube is blocking on opening a bunch of files in /sys/bus/pci/devices/*/config repeatedly | 06:01:52 |
| ersei | that doesn't explain the failure of proton, however, since the process seems to just stop | 06:03:36 |
| ersei | hm, seems like I was mistaken and the process does not stop (after disabling the steam overlay) | 06:07:04 |
| ersei | it takes a solid minute to launch (as opposed to near instant with the VK_ICD_FILENAMES set or with the nvidia gpu on) | 06:07:29 |
| ersei | seems like this is either a hardware or a nixos issue | 06:07:41 |
|  @antheww:matrix.rattlesnake.fr joined the room. | 06:41:06 |
|  hopland (meticulous montesquieu) removed their profile picture. | 22:44:19 |
| hopland (meticulous montesquieu) removed their display name IdeallyYes. | 22:46:02 |
| hopland (meticulous montesquieu) left the room. | 22:46:50 |
| 6 Feb 2024 |
| ersei | Alright, it's an issue with power management. The PCIe bridge autosuspends after a few seconds, and takes a few hundred ms to come back online. However, if you disable the autosuspend, it chews through the battery like no tomorrow. I've bumped up the autosuspend time by a few seconds, which should decrease the time it takes to scan for vulkan devices, but it's not a great solution (as it still decreases battery life) | 02:11:35 |
|  @Sweenu:matrix.org left the room. | 07:54:02 |
| @atemu12:matrix.org | ersei: Reason #2415 to not get dual-GPU laptops... | 12:12:15 |
 K900 | (*that are Nvidia) | 12:21:12 |
|  @mrtz_m:matrix.org left the room. | 19:54:44 |
| 7 Feb 2024 |
|  @jsteinke:matrix.org set a profile picture. | 01:30:10 |
|  @philipp4402:matrix.org joined the room. | 09:22:30 |
|  mrhappy200 joined the room. | 09:58:46 |
| 8 Feb 2024 |
|  @genevieve:envs.net joined the room. | 00:55:23 |
| @genevieve:envs.net | does anyone here know of a fix for certain applications using software/CPU rendering? for example, runelite always only detects llvmpipe as my GPU, and glxinfo returns OpenGL renderer string: llvmpipe (LLVM 16.0.6, 256 bits) | 00:57:18 |
| @genevieve:envs.net | I don't have anything related to amd in my nixos config, just
hardware.opengl.extraPackages = with pkgs; [
rocmPackages.rocm-runtime
rocmPackages.clr
rocmPackages.clr.icd
];
| 00:59:29 |
| @genevieve:envs.net | and I'm on mesa 23.3.3 (latest unstable) | 00:59:37 |
| @genevieve:envs.net | in general things have been kind of wonky for months - I still have to use steam -disable-cef-gpu, vkcube fails with DRI3 not available. this is definitely new and worse, though: I typically test with minetest since it's quick and easy, and that's also extremely slow and using CPU rendering, even though it used to be working fine | 01:01:22 |
| @genevieve:envs.net | oh, and I do have hardware.opengl.enable = true; | 01:12:16 |
| K900 | My guess would be mismatched Mesa versions in the closure somewhere | 06:49:34 |
|  symys joined the room. | 07:20:44 |
