Ohh, I don’t use their web based application… The compiler is open-source, so I just run it completely offline

Github repo for reference on how to install the compiler/CLI version

Big fan of Typst these days;

Markdown-like syntax, with LaTeX typesetting, great for technical documents at least

From what I’ve read Galaxy allows for downloading specific version of the games.

Might be misremembering, so take it with a pinch of salt

Edit: from https://www.reddit.com/r/linux_gaming/comments/1pxnkxj/so_for_people_who_says_we_dont_need_gog_galaxy/

It seems it’s mostly “proper cloud saving” that’s missing relative to Heroic 🤷

Surely they’ll make a native Linux client then, right, right?!

(Meanwhile; shoutout to Heroic)

Any particular reason you’re not on/tested the v590-branch? Of course it might be borked compared to 570, but might be worth a shot?

FYI:

the default, one of AMDVLK and RADV

I think that AMDVLK is shipping RADV per default now, so you might as well use RADV (for gaming, there might be other differences in OpenCL etc.), see Github/AMDVLK for more information;

In a move to streamline development and strengthen our commitment to the open-source community, AMD is unifying its Linux Vulkan driver strategy and has decided to discontinue the AMDVLK open-source project, throwing our full support behind the RADV driver as the officially supported open-source Vulkan driver for Radeon™ graphics adapters.

Why not just grab the previous PKGBUILD version that worked?

Not sure if you have the same problem or not, but I had intermittent jitter spikes (and/or complete package drops) every 60 seconds on my Realtek chipset, ran:

sudo iw dev wlan0 set power_save off

And it’s been stable since (just had to make a udev rule to make it persistent across boot)

amdvlk, any reason you’re not using the MESA driver, RADV?

… I can’t undervolt my card…

People usually use/recommend LACT for undervolting/overlocking on Linux

What is your:

1. Kernel
2. Linux-firmware
3. Mesa

Versions?

Additionally, what does the kernel logs tell you when you get a green screen?

Have you enabled Southern Islands support as a kernel parameter? Your generation of GPU was originally supported on radeon, so you need to explicitly enable SI (Southern Islands) support to use amdgpu.

See ArchWiki for more information

Pretty sure that information is stored in the driver, so you should be able to query it using monitoring software, i.e. see:

NVML-API

I know tooling like nvtop uses the API, but unsure whether it displays the maximum temperature

There is a patch-series that is now merged into 6.15-rc1 that exposes more i2c/RGB controllers on the GPU side for AMD cards, so that kernel + OpenRGB (as others have mentioned) might be a solution down the line

Patch series for reference: https://lists.freedesktop.org/archives/amd-gfx/2025-January/118399.html

In general it sounds like you want ‘tiling’. There are multiple window managers that does this, e.g. AwesomeWM, i3, Sway, River etc.

Additionally you typically have ‘tiling scripts’ that work on top of Gnome and Kwin (Plasma), however unsure what the capabilities are there.

I can atleast speak for Sway:

Here you can can move/select the current focused window relative to whatever key strokes you prefer, the defaults are using Vim-bindings, but arrow keys are also pretty common.

For grabbing a specific window (like in an ordered manner) is probably something that you would need to extend through scripting if the ‘basic’ movement isn’t enough.

Note: A tiling window manager is quite different (in usage) from a stacking one (which is what one is mostly used to) tiling capabilities/scripts

Maybe the LLMs they prompted didn’t know about the built-in SSH support, hence still recommends PuTTY? 🤔

Der8auer’s video is worth a watch, he got one of the Redditor’s card:

https://youtu.be/Ndmoi1s0ZaY

Yes, so R&D and finalizing the model weight is done on NVIDIA GPUs (I guess you need an excessive amount of VRAM).

Inference is probably gonna be offloaded to consumers in the end where the NPU is taking care of the inference cost (See Apple, Qualcomm etc)

Not the best on AI/LLM terms, but I assume that training the models was done on Nvidia, while inference (using the model/getting the data from the model) is done on Huawei chips

To add: Training the model is a huge single-cost expense, while inference is a continuous expense.

Are you by chance using an integrated GPU?

Noticed that my AMD Radeon 680M uses quite a lot of RAM as shared memory.

Using something like amdgpu_top will show how much RAM your iGPU is using, metric is ‘GTT’