You are not logged in.
Greetings,
Sorry if this question has been made 1000 times.
This is not about team red vs team green, open source drivers vs closed, etc
I have been a user of AMD for a while in Linux (I only run Arch). And it has time to upgrade. Currently while I do game (3440x1440p), my PC is not used primary for that,
I started Develop some music related plug-ins (some with neural networks) and hope to get into video editing soon.
If it wasn't for that, AMD would be an easy answer. But I do want to take (If any) advantage for Neural Networks training and also CUDA support for video editing.
Wanted to ask if for Nvidia, it is painful for kernel upgrades, compatibility issues, and what is the current status on support of Nvidia in linux for taking advantage of CUDA and AI related things? (On arch at least which is more bleeding edge?)
Whats the status of Wayland support?
Will be used with a Ryzen 9 9900X
Dont mind if for gaming is 5% slower in Linux if I will gain advantages on the other topics mentioned.
Dont want to start a debate about open source vs closed source nature of the drivers (While I do prefer AMD in that regard) , pricing, power consumption etc. Those are the 2 only cards I can get.
Thanks!
Last edited by AlmaMater (2025-08-31 18:42:12)
Offline
just don't buy these low end cards!
the nvid one comes with x8 pci-e only - if you invest in new hardware at least spent as much for a card with full 16 lanes
60-class gpus are just waste of money
Offline
Fair, but the question remains. Even if it is a 5070 against a 9070.
Offline
Wanted to ask if for Nvidia, it is painful for kernel upgrades,
It varies with each major kernel release. Some have full compatibility on release day, some need days or weeks for patches.
what is the current status on support of Nvidia in linux for taking advantage of CUDA and AI related things?
Good apart from the current lack of CUDA 13. For some AI stacks you may need a venv or namespace with older or newer libraries.
Dont want to start a debate about open source vs closed source nature of the drivers
50 series is only support by Nvidia's open kernel driver.
Offline
AlmaMater wrote:Wanted to ask if for Nvidia, it is painful for kernel upgrades,
It varies with each major kernel release. Some have full compatibility on release day, some need days or weeks for patches.
AlmaMater wrote:what is the current status on support of Nvidia in linux for taking advantage of CUDA and AI related things?
Good apart from the current lack of CUDA 13. For some AI stacks you may need a venv or namespace with older or newer libraries.
AlmaMater wrote:Dont want to start a debate about open source vs closed source nature of the drivers
50 series is only support by Nvidia's open kernel driver.
Thanks!
Offline