nvidia-gfxG06

Edit Package nvidia-gfxG06
No description set
Source Files
Filename Size Changed
README 0000000410 410 Bytes
Xwrapper 0000001173 1.15 KB
_constraints 0000000109 109 Bytes
_service 0000000609 609 Bytes
alternate-install-present 0000000258 258 Bytes
backlight.patch 0000006966 6.8 KB
check-build.sh 0000000166 166 Bytes
generate-service-file.sh 0000001414 1.38 KB
json-to-pci-id-list.py 0000002010 1.96 KB
kmp-filelist 0000001082 1.06 KB
kmp-post.sh 0000005635 5.5 KB
kmp-postun.sh 0000000783 783 Bytes
kmp-pre.sh 0000000085 85 Bytes
kmp-preun.sh 0000000581 581 Bytes
kmp-trigger.sh 0000004086 3.99 KB
kmp-triggerpostun.sh 0000000320 320 Bytes
modprobe.nvidia 0000000218 218 Bytes
modprobe.nvidia.install 0000001452 1.42 KB
my-find-supplements 0000001205 1.18 KB
n_kernel_write.patch 0000001450 1.42 KB
nvidia-gfxG06.changes 0000080154 78.3 KB
nvidia-gfxG06.rpmlintrc 0000000804 804 Bytes
nvidia-gfxG06.spec 0000012126 11.8 KB
nvidia-kmod-pci-request-regions.patch 0000000785 785 Bytes
nvidia-persistenced.tar.bz2 0000001733 1.69 KB
nvidia-settings.desktop 0000000152 152 Bytes
pci_ids-525.78.01 0000009732 9.5 KB
pci_ids-525.78.01.new 0000002267 2.21 KB
preamble 0000000883 883 Bytes
x11-video-nvidiaG06.changes 0000079293 77.4 KB
x11-video-nvidiaG06.spec 0000028778 28.1 KB
Comments 19

dziobian's avatar

The nvidia drivers fail to install with dash-sh. Can you change the scriptlets to use /bin/bash as the shell?


Stefan Dirsch's avatar

Could you be more concrete, which scriptlets?


Stefan Pavli's avatar

I recently replaced nvidia G05 with G06 but noticed a broken link chain regarding 64-bit libOpenCL.so. To solve this I had to manually add /usr/lib64/libOpenCL.so pointing to /usr/lib64/libOpenCL.so.1 /usr/lib64/libOpenCL.so.1 pointing to /usr/lib64/nvidia/libOpenCL.so.1

I suspect the existing links created by G05 (or even earlier) are removed by zypper during removal of G05. Once G06 is installed updates work fine since they don't remove the links.

Expected solution: Ensure the mentioned linkchain is not broken.


Stefan Dirsch's avatar

Well. That's not really an update. G04/G05/G06 are mutually exclusive driver packages series, which are autoselected for mutually exclusive nVidia GPUs. Unfortunately it's not supported to update from G0<n> to G0<n+1> driver package series. I think I've tried supporting this, but I failed for several reasons. In this case the update-alternatives command in %preun of G05 removes the symlink, which has been created by the update-alternatives command in %post of G06 (Yes! %post of new packges rungs before %preun of old package. That's RPM ...). So please "update" here in two steps. Uninstall G0<n>, then install G0<n+1>.


Stefan Pavli's avatar

Thanks. Have 1 machine left that is not yet updated. Will try your suggestion.

BTW: My GPUs are GTX 1050 and GTX 1080 and are all supported by G05 as well as G06.


Milan Savić's avatar

Would it be possible to update the driver from 515.76 to the 520.56.06 feature branch due to the rather critical error with 30 series cards.

It is a rather big problem because the previous versions don't work with the kernel 6.0.


Stefan Dirsch's avatar

I'm not happy with switching to a pre-production driver. Never have done that. We can talk about using the beta version of production driver though. But this won't happen tomorrow. I'm not allowed talking the release schedule for these ...

I don't think anybody forces you to use Kernel 6.0 now. Also G05 packages should work with Kernel 6.0. At least I fixed the build. And RTX 30XX are still supported by G05 according to NVIDIA's documentation. Hmm ... maybe not all of them ...

Or do you mean previous 515.xx versions?



Milan Savić's avatar

I don't have a problem with locking 5.19 kernel and 515.65 driver, but any new Tumbleweed installation will use a 6.0 kernel and 515.76 driver so the problem will happen there.

I'm not sure about G05 package, last time I used it it didn't work for 30 series, but that could be a solution for some users.

PS. Is it possible to make the compilation part of the driver parallel? Right now it takes a long time (minutes) for the RPM package to compile while the NVidia binary package takes a couple of seconds.

I've noticed that if I add the following before zypper in x11-video-nvidiaG0* it actually uses parallel compilation:

export CONCURRENCY_LEVEL=`nproc` && \

export JOBS=${CONCURRENCY_LEVEL} && \

export __JOBS=${JOBS} && \

export MAKEFLAGS="-j ${JOBS}" && \

zypper in -f x11-video-nvidiaG06


Stefan Dirsch's avatar

I've enabled parallel build now in kmp-post.sh/kmp-trigger.sh now.. I hope I didn't break anything with that change. Results are mixed. I couldn't see improvements in numbers (8 CPUs but no SSD).


Milan Savić's avatar

I can't test it (due to above problem) but on arch they seem to do a sed replacement on the dkms.conf.

https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=nvidia-510xx-utils#n42

https://aur.archlinux.org/cgit/aur.git/tree/110-nvidia-change-dkms-conf.patch?h=nvidia-beta-dkms#n14

https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=nvidia-390xx-utils#n50

When I used arch I didn't have any problems related to it (ignoring other arch exclusive problems). But it's not really a big deal, just that new opensuse users could be confused why nvidia rpm installation is slow, while binary install is fast.


Stefan Dirsch's avatar

Ok. Implemented this now back to G04 and tested it on a Kepler with G05. Appears to work fine.


Alexander Ahjolinna's avatar

I dont see why there cant be a "production branch" or beta release (or whatever) separately, especially for those tumbleweed users would want to have the latest stuff , and at least those who game and want to test the latest fixes/features

I understand having a "production branch" release, but it would be nice to have a other one for those who want the latest stuff but don't want to make or add yet another repo to their system (and hope that wont break stuff)


Stefan Dirsch's avatar

The repository on NVIDIA's server is no regular openSUSE repository, within we can have various versions of packages available. Even if we could this would mean it would contain driver versions with security issues. Not sure if we really want this ...


Alexander Ahjolinna's avatar

not sure what you mean by "security issues" ...anyway just saying what some (tumbleweed) users would like to have...especially gaming community, some of them even would like to test beta drivers for the latest features/fixes

.... if those security issues are such a issue, then why not just provide the .spec etc. for those people who want the latest (beta?) drivers and they could build on their own repos (or locally)


Stefan Dirsch's avatar

It's probably not widely known but NVIDIA releases quarterly updates, which include also fixed security issues (back to 390.xx at the moment). So it's not recommended to use older driver versions.

Feel free to fork this package and prepare it for the pre-production driver. It's no rocket science to do so. I don't have the time to maintain the pre-production driver packaging in addition.


David Šmíd's avatar

The package x11-video-nvidiaG06 contains a SystemD service /usr/lib/systemd/system/nvidia-powerd.service which is supposed to start /usr/bin/nvidia-powerd , however this binary is missing in the package.


Stefan Dirsch's avatar

Thanks for the report! Apparently nvidia-powerd and the service file has been removed silently with 460.32.03 (NVIDIA's shell script archive) in G05. Later G06 was forked from G05, but 515.xx.yy apparently silently reintroduced nvidia-powerd and the service file. The latter was installed again via glob pattern, but the binary was missing now. :-( I will add it again.


Stefan Dirsch's avatar

@davidsmid done. Will be included with the next driver packages update


Mon Oct 17 09:51:41 UTC 2022 - Stefan Dirsch sndirsch@suse.com

openSUSE Build Service is sponsored by