You are not logged in.
According to this I've set things up to use the Nvidia card as my primary GPU. Is there still any need or benefit to install the mesa package?
Last edited by Ridwan Rawriet (2022-01-28 11:59:30)
Offline
I asked google if nvidia uses mesa and it said that the proprietary drivers replace it
Offline
nvidia-utils depends on libglvnd which depends on mesa.
Offline
I see, installing nvidia-utils automatically pulls libglvnd and mesa. I guess the nvidia package blacklists nouveau that mesa contains. Maybe unnecessary but how can I assure that? And can I ensure that VDPAU is being used not VA-API and vulkan, openGL are all using nvidia-utils?
Offline
libglvnd will ensure that, that's the point and purpose of it. Unless you actively manipulate your library path it will always end up on the nvidia driver as it is the active GL renderer. VDPAU and VAAPI have no relation to that in any case,so you might have to change things here, see https://wiki.archlinux.org/title/Hardwa … figuration for the relevant config options , use the relevant testing utils e.g. glxinfo/vulkaninfo/vdpauinfo etc to verify.
It is absolutely necessary you have the mesa package as well, since you still want acceleration on the intel card, albeit solely for the purpose of displaying the rendered image of the nvidia card.
Last edited by V1del (2021-12-12 20:27:02)
Offline
I see, the intel card is always used to output to my screen but the rendering is done by the nVidia card. So I cannot abandon it.
What will setting the environment variable LIBVA_DRIVER_NAME=vdpau do? Since proprietary nvidia drivers don't support VA-API?
Thanks for your kind donation of time and knowledge.
Last edited by Ridwan Rawriet (2021-12-13 07:10:38)
Offline
There is a wrapper library in libva-vdpau-driver that wraps VAAPI calls and passes them to VDPAU. Afaik not really actively maintained and the CPU overhead from doing that translation might well not really make it worth it as opposed to normal software decoding
Offline