Update 41, gnome, nvidia: GPU no longer available in Cemu
I was playing through Wind Waker HD with no issues at all at 30 FPS (the game is capped), but after my system updated, the game was running at 5 FPS. I went into Cemu's settings and I saw there's an option called "Graphics Device", where the only available option was "llvmpipe (LLVM 18.1.8, 256 bits)".
After I
rpm-ostree rollback and systemctl reboot I'm able to select my RTX 4070 Super as a "Graphics Device" and performance is back to normal.
I have toggled auto-updates off for the moment. Here's information about my system, in case it's helpful.
Image where everything worked:
● ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-gnome-nvidia:stable
Digest: sha256:28e9785f1f90699d38063655319286a01669aacdbb7ed75a90a84f78c4944799
Version: 40.20241020 (2024-10-21T05:06:50Z)
LayeredPackages: sunshine
Image where things started not working:
ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-gnome-nvidia:stable
Digest: sha256:6f541bb46616bb5b1414a6d174cf63c40e99016aa2cfab30ea9cd0da1713e497
Version: 41.20241029.1 (2024-10-29T15:38:34Z)
LayeredPackages: sunshine
Fastfetch:
bazzite-gnome-nvidia:stable
Bazzite 40 (FROM Fedora Silverblue)
Linux 6.9.12-210.fsync.fc40.x86_64
X470 AORUS ULTRA GAMING
AMD Ryzen 7 2700X (16) @ 3.70 GHz
NVIDIA GeForce RTX 4070 SUPER
3.56 GiB / 15.55 GiB (23%)20 Replies
I think I’m also having the same issue. PCSX2 is almost unplayable now, but was just fine before.
It's really similar for me. Robocop Rouge City runs less than half fps, furmark doesn't start, basemark only finds the llvm, not the GPU since update. Gnome, ryzen 5 3600, nvidia 1080gtx. I updated like 2 days ago, I'm not at home but will check the version if it helps.
if you have a turing card or newer (GTX 16xx or newer) then you should rebase to the
nvidia-open image
if you got an older card then you stay on the nvidia image
gnome image: rpm-ostree rebase ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-gnome-nvidia-open:stable
kde image: rpm-ostree rebase ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-nvidia-open:stable
if you have a GTX 10xx card do not rebase to nvidia-open, your card will not work thenI read the news about bazzite 41 but don't remember reading about this further branching of nvidia images. Where can I get some info about the reasons? Also, my gtx1080 is not working with bazzite 41. I had bazzite 40 and simply ujust update-ed it. I have auto update turned off. What's the default path in this case for the update? If you help me how to check it, I'll do that after I get home
the reason is nvidia is only going to focus on nvidia-open for newer cards. new cards will not support the legacy closed source kernel module so we need 2 images
1 for old cards and 1 for new cards
for the gtx1080 make sure youre not on the nvidia-open image
lots of people fail at reading when on the download page and do not read that the nvidia option is for newer cards and that there is a legacy nvidia option for 10 series
other than that i know little about nvidia as i havent used nvidia in linux for almost a decade
When I installed bazzite around a month ago, there was only one gnome-nvidia image (for gtx 10xx), I installed that one and haven't updated it since. That's my first system update. How can I check which version is installed?
go to Terminal and type
rpm-ostree status
you should see 2 sections, one with dot is the active version while the other is the backupYes, I saw that before but haven't realized it will give me the info. So, there, the image name/source shouldn't include the nvidia-open, right? If I remember correctly, my branch is something like bazzite-gnome-nvidia
yes, as long as you don't rebase to any image, it should stay on
bazzite-gnome-nvidiaI'll check it, but if that's the case, my problem stays. For some reason, the new image has serious issues with my GPU while the bazzite 40 image works flawless
These are the images on my desktop:
Can I do anything to figure out why the GPU is not working well on the new one, or should I just roll back and hope that some new version performs better?
I don't know it helps at all, but here is the output from
nvidia-smi
i would say rollback and pin the old image or rebase to the 20241020 image and pin it and wait
nvidia release only a handful of driver updates per year for linux
Oh, some new info. What is pinning an image? It sounds like something I was already thinking about. Is it like not letting it go away, no matter how many images I install?
rpm-ostree status -v will show you the index for each deployment on your machine and other info
sudo ostree admin pin INDEXNUMBERHERE will pin that deployment so it will not get deleted (as ostree by default only keeps 2 deployments with 1 transitional deployment for updates)
to unpin you do
sudo ostree admin pin --unpin INDEXNUMBERHERE
pinning a image deployment will add an entry for it into grub when it would otherwise have been deleted so you can still boot it.
if you didnt pin it, it would be deleted from the machine.awesome, thanks! That's exactly what I need now
if you rebase to the image of the specific date of the working image, it will be the default boot option
a pinned image will always be option 3 or more
also keep in mind you only have space in /boot for a few pinned images so dont go overboard 😛
For now, I only need this one to not disappear. This is my preciouss image, because it works flawlessly
i always recommend having 1 "golden" image pinned for situations like this
and you just rotate it out when you can
yes, I just rolled back and pinning the image now. Thank you for the tip!
The index number is the Digest or the Commit number?
oh, I see now. Index: 0 and 1
yep
you will still get updates when on the pinned image
rebasing to the date will instead lock you to that image, until you rebase back to stable
How can I list the pinned images?
I seeee. rpm-ostree status adds a pinned message