The answer to video title - yes, objectively. As someone who prefers VSync, it caused me a lot of trouble on windows. Because for some reason whenever there's an artificial limit to framerate such as VSync, the GPU's frequency stays low. Seemingly because AMD driver on windows tries to "guess" whenever you're gaming or not. So when it sees low GPU utilization (such as with a framerate limiter) it doesn't let card accelerate which brought me stutter on windows. Uncapping FPS in the game fixes this issue. But it's just nonsense in my opinion. Linux has no issues with VSync and utilizes GPU properly.
Thanks - very interesting. I don't do much gaming, mostly just coding. I have two AMD systems (GPU and CPU), one system with an Intel CPU and AMD GPU, and one system with an Intel CPU and Nvidia GPU. I don't really have a preference at this stage, haven't had any issues with any of my systems. I have to use the Nvidia GPU for CUDA projects, but I'm now looking into AMD alternatives. I think video hardware encoding might be easier on Nvidia too.
Having owned an Nvidia GPU before really getting into Linux on my desktop, the issues I've had have been less about drivers and making the GPU work under Linux, and more about independent projects making it clear that they want nothing to do with Nvidia. I'm currently using Sway on top of wlroots, which seem to be particularly anti-Nvidia. The wlroots package has a 2 line Nvidia patched version on the AUR, that seems to intentionally not be accepted upstream to prevent horrible eye-straining flickering, not to mention being forced to use "--unsupported-gpu" and being told "not to submit bug reports" if I experience issues like this. A wayland VNC host project I wanted to use is impossible, solely because the maintainer doesn't want to support EGL, not because there's a flaw with Nvidia.
I've seen more than a few projects that won't support Nvidia GPUs as it ends up being a maintenance burden on them, and especially when the maintainers themselves don't have the hardware. I don't know if anyone with an Nvidia card has actively approached them to suggest working together for long-term maintenance. I recall GNOME and KDE both having to put in significant work to keep two versions of their code split for Mesa vs Nvidia. I see it in other areas of open-source development, like I believe it's Handbrake that won't support VAAPI for video encoding.
I agree with this to an extent - Gnome and KDE having to make major changes to make things work across both Nvidia and AMD sounds like something of a nightmare for smaller projects - and if one man projects choose not to support Nvidia for the sheer effort of doing so, I don't find it offensive. What I do find offensive is the current wlroots situation. Currently I am having to manually apply three separate very short patches that simply add dmabuff capture and changes a flush to a finish to prevent major flickering. Testing this patched version across a borrowed AMD card and Intel integrated show no difference in the resulting image nor benchmarks, leading me to believe that for very little reason, these issues being brought up in the wlroots/Sway projects are being intentionally ignored, despite simple fixes existing. While I can understand the nvidia drivers may be a hassle to support, when the bare essentials are so simple to fix and have already been available, it's hard to understand why it would be ignored in the way it is. It's also encouraging, however, that I was still able to get it to work with community support.
AMD CPUs Nvidia GPUs been using and playing on Linux since Ubuntu 14.04. Since 75-80% of GPUs are Nvidia it's good for all of us that love Linux to figure them all out.
If you don't care about cuda programming, ai or video editing/3d modeling than just buy AMD. Simple as that. Wayland is trash. Unoptimised piece of crap.
I do all of this and still use AMD on Wayland. I use Stable diffusion and Large Language model AI and do 3D modelling then render 4K60fps video with Blender cycle. People may believe whatever they want but all this work pretty good on AMD. The AI part is actually surprising because it is already as fast as Nvidia. The ROCm thing is pretty functional, most of the main AI project have ported their Cuda code to HIP already.
@@MumblingHugo Yes Davinci need ROCm/opencl support installed. And AMF need amdvlk. So there is couple of thing to install for sure. If I remember correctly not all IGPU are supported. So Davinci may be troublesome on a laptop.
