You don’t want to, though. They’re horrible. There’s an insane amount of effort that would be required to reverse-engineer drivers since Nvidia is at best negligent. AMD and Intel are much better about OSS.
I’m not fully a penguin, but getting there. Saw the memes, experienced it first hand in one case and was plug and play in another. It’s luck of the draw.
There was one for Windows too, but most people don’t remember the hundreds of hours of learning that they’ve done to become competent users of Windows.
Just jump in, don’t dual boot. Having no option of giving up and booting Windows makes you motivated to learn how to use Linux.
There’s a community of people who will help (while also sometimes being insufferable assholes) and the skills you learn will be more durable. You’re not going to see Linux 11 come along and mandate that you buy a new computer or anything.
Oh I’ve loved it so far. And you’re right on the “what you learn is more useful”. Like I’d done a fair amount of hobby/work prototype stuff on rasbian, and eventually went “man, it’d be great if this but more horsepower” and wound up Debian.
Anyway, my point is despite doing a fair amount of coding, and circuit level electronics including troubleshooting comms and all the fun things like race conditions that go into that, I had zero idea how a computer was actually arranged. Troubleshooting Debian helped me with that and is infinitely transferable as opposed to being a tip and trick with windows.
But my original comment was just about Nvidia cards. I’ve had some I just slot in and they work, and some I have to spend an afternoon troubleshooting. Still reinforces your point though, troubleshooting it the first time was how I learned how things actually get displayed.
they are at a point where it’s not even really limited by reverse engineering, but that only the nvidia-signed drivers can increase the gpu’s frequencies to anything near performant.
That is no longer the case. The Nvidia drivers for Linux are pretty decent, these days. They’re still closed source, so if that’s a deal breaker for you, you’ll need to buy an AMD GPU.
The problem is not that they are bad, is that if someone makes a project that depends on the specific drivers then it will work much worse if the drivers are closed source. Wayland was unusable with nvidia drivers until recently.
NVK is very slowly getting there, from what I’ve read. if I remember correctly, it’s still gives horrible performance (about 50%-ish of the closed source ones, I think?), but it’s still miles better than “you’re really better off using your integrated GPU” that noveau offered for ages.
Fair enough. I wish them well in the effort. It would be nice if Nvidia threw them a bone, though, what with all the AI money and their GPUs being used in so many Linux supercomputers and servers.
You don’t want to, though. They’re horrible. There’s an insane amount of effort that would be required to reverse-engineer drivers since Nvidia is at best negligent. AMD and Intel are much better about OSS.
been plug and play for me using Nvidia + Linux for years now. Just upgraded to a 5070ti, literally was take out old, put in the new.
I’m not fully a penguin, but getting there. Saw the memes, experienced it first hand in one case and was plug and play in another. It’s luck of the draw.
There’s a learning curve, sure.
There was one for Windows too, but most people don’t remember the hundreds of hours of learning that they’ve done to become competent users of Windows.
Just jump in, don’t dual boot. Having no option of giving up and booting Windows makes you motivated to learn how to use Linux.
There’s a community of people who will help (while also sometimes being insufferable assholes) and the skills you learn will be more durable. You’re not going to see Linux 11 come along and mandate that you buy a new computer or anything.
Oh I’ve loved it so far. And you’re right on the “what you learn is more useful”. Like I’d done a fair amount of hobby/work prototype stuff on rasbian, and eventually went “man, it’d be great if this but more horsepower” and wound up Debian.
Anyway, my point is despite doing a fair amount of coding, and circuit level electronics including troubleshooting comms and all the fun things like race conditions that go into that, I had zero idea how a computer was actually arranged. Troubleshooting Debian helped me with that and is infinitely transferable as opposed to being a tip and trick with windows.
But my original comment was just about Nvidia cards. I’ve had some I just slot in and they work, and some I have to spend an afternoon troubleshooting. Still reinforces your point though, troubleshooting it the first time was how I learned how things actually get displayed.
It’s not luck it’s pretty well defined what works
they are at a point where it’s not even really limited by reverse engineering, but that only the nvidia-signed drivers can increase the gpu’s frequencies to anything near performant.
I know. The best you can do in Linux is not use nvidia.
That is no longer the case. The Nvidia drivers for Linux are pretty decent, these days. They’re still closed source, so if that’s a deal breaker for you, you’ll need to buy an AMD GPU.
I’m not sure if the closed-source drivers have social media garbage on them at the moment, but I’m very sure that I don’t trust Nvidia not to add it.
This is quite frankly nonsense
when was the last time you used windows with nvidia graphics?
The problem is not that they are bad, is that if someone makes a project that depends on the specific drivers then it will work much worse if the drivers are closed source. Wayland was unusable with nvidia drivers until recently.
NVK is very slowly getting there, from what I’ve read. if I remember correctly, it’s still gives horrible performance (about 50%-ish of the closed source ones, I think?), but it’s still miles better than “you’re really better off using your integrated GPU” that noveau offered for ages.
Fair enough. I wish them well in the effort. It would be nice if Nvidia threw them a bone, though, what with all the AI money and their GPUs being used in so many Linux supercomputers and servers.
There’s a project working on making CUDA work on all (read: AMD) graphics cards. It’s alpha-level, but the progress makes it look promising.
https://www.tomshardware.com/software/a-project-to-bring-cuda-to-non-nvidia-gpus-is-making-major-progress-zluda-update-now-has-two-full-time-developers-working-on-32-bit-physx-support-and-llms-amongst-other-things
e: Tom’s Hardware links are half the size of the article 😂