Created attachment 14817 [details]
files list

Note that starting from this version also the dkms-nvidia-current-open-550.142-1.mga9.nonfree is provided as the alternative opensourced nvidia kernel modules (this package it's not automatically retrieved from drakx11). Note that, while the nvidia-current package works for cards of series GTX 745 and beyond, these modules, as we already seen for nvidia-newfeature, works only for cards of series Turing and beyond.

Text from Comment 2 added to https://wiki.mageia.org/en/Setup_the_graphical_server#Nvidia_proprietary_drivers

We should also note it in release notes for mga10.

CC: (none) => fri
Keywords: (none) => FOR_RELEASENOTES10

MGA9-64 Plasma. 

$ inxi -MCG
Machine:
  Type: Desktop Mobo: ASUSTeK model: PRIME Q270M-C v: Rev X.0x
    serial: <superuser required> UEFI: American Megatrends v: 2201
    date: 12/21/2023
CPU:
  Info: quad core model: Intel Core i5-7500 bits: 64 type: MCP cache:
    L2: 1024 KiB
  Speed (MHz): avg: 800 min/max: 800/3800 cores: 1: 800 2: 800 3: 800 4: 800
Graphics:
  Device-1: NVIDIA GM107GL [Quadro K620] driver: nvidia v: 550.142
  Display: x11 server: X.org v: 1.21.1.8 with: Xwayland v: 22.1.9 driver: X:
    loaded: nvidia,v4l gpu: nvidia,nvidia-nvswitch resolution: 1920x1080~60Hz
  API: EGL v: 1.5 drivers: nvidia,swrast
    platforms: gbm,x11,surfaceless,device
  API: OpenGL v: 4.6.0 vendor: nvidia v: 550.142
    renderer: Quadro K620/PCIe/SSE2
  API: Vulkan v: 1.3.231 drivers: nvidia,llvmpipe surfaces: xcb,xlib

First test is a simple update of the existing packages, as installed by XFdrake.
No installation issues, and no issues in a very brief test after the reboot.

More to come later today.

CC: (none) => andrewsfarm

I was going to use XFdrake from MCC to switch to the nvidia470 driver, but as with the newfeature driver under test it says dkms -nvidia-current-open can't be selected because there's a conflict with dkms-nvidia470. 

I told it to continue anyway, rather than abort like I did when testing the newfeature driver. It seemed to proceed normally, but when finished it told me it couldn't install the x11 driver for nividia470. But, when I checked with drakrpm, it said the x11 driver WAS installed.

I'm confused, and the way things are at the moment, I would expect our users to be confused, as well.

Time to try rebooting, supposedly using nvidia470...

Installed and tested without issues.

Tested:
- OpenGL: glmark2, glinfo;
- Vulkan: vkcube, vulkaninfo;
- OpenCL: clinfo;
- VDPAU: vdpauinfo, vlc, smplayer;
- WebGL: Firefox, Chromium;
- Plasma DE;
All OK.


System: Mageia 9, x86_64, Plasma DE, Xorg, Intel Core i7-8750H, Quadro P600 Mobile using nvidia-current proprietary driver.



$ uname -a
Linux saturno 6.6.65-desktop-2.mga9 #1 SMP PREEMPT_DYNAMIC Thu Dec 12 12:42:26 UTC 2024 x86_64 GNU/Linux
$ cat /proc/driver/nvidia/version
NVRM version: NVIDIA UNIX x86_64 Kernel Module  550.142  Wed Dec 11 05:15:39 UTC 2024
GCC version:  gcc version 12.3.0 (Mageia 12.3.0-3.mga9) 
$ rpm -qa | grep -P 'nvidia.*550.142' | sort
dkms-nvidia-current-550.142-1.mga9.nonfree
nvidia-current-cuda-opencl-550.142-1.mga9.nonfree
nvidia-current-devel-550.142-1.mga9.nonfree
nvidia-current-doc-html-550.142-1.mga9.nonfree
nvidia-current-utils-550.142-1.mga9.nonfree
x11-driver-video-nvidia-current-550.142-1.mga9.nonfree
$ inxi -SMCGN
System:
  Host: saturno Kernel: 6.6.65-desktop-2.mga9 arch: x86_64 bits: 64
  Desktop: KDE Plasma v: 5.27.10 Distro: Mageia 9
Machine:
  Type: Laptop System: FUJITSU product: CELSIUS H780 v: N/A
    serial: <superuser required>
  Mobo: FUJITSU model: FJNBB5D v: 757383-01R9800096
    serial: <superuser required> UEFI: FUJITSU // Insyde v: Version 1.21
    date: 09/14/2022
CPU:
  Info: 6-core model: Intel Core i7-8750H bits: 64 type: MT MCP cache:
    L2: 1.5 MiB
  Speed (MHz): avg: 800 min/max: 800/4100 cores: 1: 800 2: 800 3: 800 4: 800
    5: 800 6: 800 7: 800 8: 800 9: 800 10: 800 11: 800 12: 800
Graphics:
  Device-1: NVIDIA GP107GLM [Quadro P600 Mobile] driver: nvidia v: 550.142
  Device-2: Chicony Integrated Camera driver: uvcvideo type: USB
  Display: x11 server: X.org v: 1.21.1.8 with: Xwayland v: 22.1.9 driver: X:
    loaded: nvidia,v4l gpu: nvidia,nvidia-nvswitch resolution: 1920x1080~60Hz
  API: EGL v: 1.5 drivers: nvidia,swrast
    platforms: gbm,x11,surfaceless,device
  API: OpenGL v: 4.6.0 vendor: nvidia v: 550.142
    renderer: Quadro P600/PCIe/SSE2
  API: Vulkan v: 1.3.231 drivers: nvidia,llvmpipe surfaces: xcb,xlib
Network:
  Device-1: Intel Cannon Lake PCH CNVi WiFi driver: iwlwifi
  Device-2: Intel Ethernet I219-LM driver: e1000e

(In reply to Thomas Andrews from comment #5)
> I was going to use XFdrake from MCC to switch to the nvidia470 driver, but
> as with the newfeature driver under test it says dkms -nvidia-current-open
> can't be selected because there's a conflict with dkms-nvidia470. 
> 
> I told it to continue anyway, rather than abort like I did when testing the
> newfeature driver. It seemed to proceed normally, but when finished it told
> me it couldn't install the x11 driver for nividia470. But, when I checked
> with drakrpm, it said the x11 driver WAS installed.
> 
> I'm confused, and the way things are at the moment, I would expect our users
> to be confused, as well.
> 
> Time to try rebooting, supposedly using nvidia470...

have you updated the meta-task and mageia-repos from updates_testing too in this subject? This should resolve the preferred package when two alternatives (-open and plain) arise.

I did not. They weren't in the file list.

However, the reboot with nvidia470 was without incident. From there I used MCC once more to switch to nvidia-current, and that proceeded normally, with no error messages. A reboot from using that was also without incident.

(In reply to Giuseppe Ghibò from comment #7)
> 
> have you updated the meta-task and mageia-repos from updates_testing too in
> this subject? This should resolve the preferred package when two
> alternatives (-open and plain) arise.

I used qarepo to get meta-task* and mageia-repos* as well as the other files listed for this bug, and installed the meta-task and mageia-repos updates without incident.

I rebooted once more, just to be sure, and then once again attempted to use MCC to switch to the nvidia470 driver, and once again it told me that dkms-nvidia-current-open conflicts with dkms-nvidia470.

(In reply to Thomas Andrews from comment #9)
> (In reply to Giuseppe Ghibò from comment #7)
> > 
> > have you updated the meta-task and mageia-repos from updates_testing too in
> > this subject? This should resolve the preferred package when two
> > alternatives (-open and plain) arise.
> 
> I used qarepo to get meta-task* and mageia-repos* as well as the other files
> listed for this bug, and installed the meta-task and mageia-repos updates
> without incident.

I added also meta-task and mageia-repos here. Files list need to be updated.

> 
> I rebooted once more, just to be sure, and then once again attempted to use
> MCC to switch to the nvidia470 driver, and once again it told me that
> dkms-nvidia-current-open conflicts with dkms-nvidia470.

I've updated the release to nvidia-current-550.142-2.mga9.nonfree, and also a newer nvidia470-470.256.02-2.mga9 with an extra explicit Conflicts with dkms-nvidia-current-open. Ditto for nvidia-newfeature bumped at nvidia-newfeature-565.77-4.mga10.

There will be a newer ldetect-lst-0.59 with a minor adding about an RTX 3050 mobile PCI-id that was added to newer 550.142, package actually in the preparing.

(In reply to Giuseppe Ghibò from comment #10)

> There will be a newer ldetect-lst-0.59 with a minor adding about an RTX 3050
> mobile PCI-id that was added to newer 550.142, package actually in the
> preparing.

I mean ldetect-lst-0.6.59-1.mga9.

MGA9-64, ‎AMD Ryzen 5 2600, Nvidia 1650 super, GNOME

The following 6 packages are going to be installed:

- dkms-nvidia-current-550.142-1.mga9.nonfree.x86_64
- meta-task-9-3.mga9.noarch
- nvidia-current-cuda-opencl-550.142-1.mga9.nonfree.x86_64
- nvidia-current-doc-html-550.142-1.mga9.nonfree.x86_64
- nvidia-current-utils-550.142-1.mga9.nonfree.x86_64
- x11-driver-video-nvidia-current-550.142-1.mga9.nonfree.x86_64

141KB of additional disk space will be used.

---- rebooted

$ nvidia-smi
Tue Dec 24 07:55:53 2024       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.142                Driver Version: 550.142        CUDA Version: 12.4     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce GTX 1650 ...    Off |   00000000:07:00.0  On |                  N/A |
| 35%   31C    P8             11W /  100W |     273MiB /   4096MiB |      2%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+


browser working
calc working


works for me

---

per comment 11 - will wait for minor update before moving forward.

CC: (none) => brtians1

MGA9-64, Plasma, Ryzen 5600, Nvidia 1050

installed open versions



--rebooted

That made a mess and was not compatible with my 1050GT card.  I had to go to an older kernel and restore nouveau so I could uninstall.

We need to update our choices in the XFdrake utility.  Who supports that?


-- after recovering with Nouveau I went back to install 550.142 proprietary.  They work fine.  Did run into this mapping issue for dependencies maybe.

Nvidia-current-all (64 bit) picked the following:
To satisfy dependencies, the following package(s) also need to be installed:

- libbrotlicommon1-1.0.9-5.mga9.i586
- libbrotlidec1-1.0.9-5.mga9.i586
- libbz2_1-1.0.8-5.mga9.i586
- libcom_err2-1.47.0-2.mga9.i586
- libcurl4-7.88.1-4.5.mga9.i586
- libdri-drivers-24.2.8-1.mga9.tainted.i586
- libdrm2-2.4.123-1.mga9.i586
- libdrm_amdgpu1-2.4.123-1.mga9.i586
- libdrm_intel1-2.4.123-1.mga9.i586
- libdrm_radeon1-2.4.123-1.mga9.i586
- libedit0-3.1-0.20221030.1.mga9.i586
- libegl1-1.6.0-1.mga9.i586
- libelfutils1-0.189-1.mga9.i586
- libexpat1-2.6.4-1.mga9.i586
- libffi8-3.4.4-1.mga9.i586
- libgbm1-24.2.8-1.mga9.tainted.i586
- libgcrypt20-1.10.2-2.mga9.i586
- libgl1-1.6.0-1.mga9.i586
- libglapi0-24.2.8-1.mga9.tainted.i586
- libgldispatch0-1.6.0-1.mga9.i586
- libglesv1_cm1-1.6.0-1.mga9.i586
- libglesv22-1.6.0-1.mga9.i586
- libglx0-1.6.0-1.mga9.i586
- libgpg-error0-1.47-1.mga9.i586
- libidn2_0-2.3.4-3.1.mga9.i586
- libkeyutils1-1.6.3-2.1.mga9.i586
- libkrb53-1.20.1-1.3.mga9.i586
- libldap2.5_0-2.5.14-1.mga9.i586
- libllvm15.0-15.0.6-1.mga9.i586
- liblm_sensors5-3.6.0-4.mga9.i586
- liblzma5-5.4.3-1.mga9.i586
- libmesaegl1-24.2.8-1.mga9.tainted.i586
- libmesagl1-24.2.8-1.mga9.tainted.i586
- libncurses6-6.3-20221203.2.1.mga9.i586
- libnghttp2_14-1.61.0-1.mga9.i586
- libnvidia-egl-wayland1-1.1.11-1.mga9.i586
- libopencl1-2.3.1-2.mga9.i586
- libopengl0-1.6.0-1.mga9.i586
- libopenssl3-3.0.15-1.1.mga9.i586
- libpciaccess0-0.17-1.mga9.i586
- libpsl5-0.21.2-2.mga9.i586
- libsasl2_3-2.1.27-7.mga9.i586
- libssh4-0.10.6-1.mga9.i586
- libunistring5-1.1-1.mga9.i586
- libverto1-0.3.2-4.mga9.i586
- libwayland-client0-1.21.0-1.mga9.i586
- libwayland-server0-1.21.0-1.mga9.i586
- libx11-xcb1-1.8.6-1.1.mga9.i586
- libx11_6-1.8.6-1.1.mga9.i586
- libxau6-1.0.11-1.mga9.i586
- libxcb-dri2_0-1.15-2.mga9.i586
- libxcb-dri3_0-1.15-2.mga9.i586
- libxcb-glx0-1.15-2.mga9.i586
- libxcb-present0-1.15-2.mga9.i586
- libxcb-randr0-1.15-2.mga9.i586
- libxcb-shm0-1.15-2.mga9.i586
- libxcb-sync1-1.15-2.mga9.i586
- libxcb-xfixes0-1.15-2.mga9.i586
- libxcb1-1.15-2.mga9.i586
- libxdmcp6-1.1.4-1.mga9.i586
- libxext6-1.3.5-1.mga9.i586
- libxfixes3-6.0.0-2.mga9.i586
- libxshmfence1-1.3.2-1.mga9.i586
- libxxf86vm1-1.1.5-1.mga9.i586
- libzlib1-1.2.13-1.2.mga9.i586
- libzstd1-1.5.5-1.mga9.i586
- nvidia-current-lib32-550.142-2.mga9.nonfree.x86_64

248MB of additional disk space will be used.

That doesn't make sense to me.

(In reply to Brian Rockwell from comment #13)

> MGA9-64, Plasma, Ryzen 5600, Nvidia 1050
> 
> installed open versions
> 

The NVidia GTX 1050 is of architecture Pascal, so it's not compatibile with the dkms-nvidia-current-open or dkms-nvidia-newfeature-open, because the minimal requirement for -open is of architecture "Turing". So while the GTX 1050 is compatible with 5 series of drivers (nouveau, modesetting, nvidia470, nvidia-current, and nvidia-newfeature), it is not regarding the -open modules variant.

> 
> 
> --rebooted
> 
> That made a mess and was not compatible with my 1050GT card.  I had to go to
> an older kernel and restore nouveau so I could uninstall.
> 
> We need to update our choices in the XFdrake utility.  Who supports that?

The -open driver modules are not automatically installed by XFdrake in any choice. The "alternative" installation is left manually to user.

> 
> 
> -- after recovering with Nouveau I went back to install 550.142 proprietary.
> They work fine.  Did run into this mapping issue for dependencies maybe.
> 
> Nvidia-current-all (64 bit) picked the following:
> To satisfy dependencies, the following package(s) also need to be installed:
> 
> - libbrotlicommon1-1.0.9-5.mga9.i586
> - libbrotlidec1-1.0.9-5.mga9.i586
> - libbz2_1-1.0.8-5.mga9.i586
> - libcom_err2-1.47.0-2.mga9.i586
> - libcurl4-7.88.1-4.5.mga9.i586
> - libdri-drivers-24.2.8-1.mga9.tainted.i586
> - libdrm2-2.4.123-1.mga9.i586
> - libdrm_amdgpu1-2.4.123-1.mga9.i586
> - libdrm_intel1-2.4.123-1.mga9.i586
> - libdrm_radeon1-2.4.123-1.mga9.i586
> - libedit0-3.1-0.20221030.1.mga9.i586
> - libegl1-1.6.0-1.mga9.i586
> - libelfutils1-0.189-1.mga9.i586
> - libexpat1-2.6.4-1.mga9.i586
> - libffi8-3.4.4-1.mga9.i586
> - libgbm1-24.2.8-1.mga9.tainted.i586
> - libgcrypt20-1.10.2-2.mga9.i586
> - libgl1-1.6.0-1.mga9.i586
> - libglapi0-24.2.8-1.mga9.tainted.i586
> - libgldispatch0-1.6.0-1.mga9.i586
> - libglesv1_cm1-1.6.0-1.mga9.i586
> - libglesv22-1.6.0-1.mga9.i586
> - libglx0-1.6.0-1.mga9.i586
> - libgpg-error0-1.47-1.mga9.i586
> - libidn2_0-2.3.4-3.1.mga9.i586
> - libkeyutils1-1.6.3-2.1.mga9.i586
> - libkrb53-1.20.1-1.3.mga9.i586
> - libldap2.5_0-2.5.14-1.mga9.i586
> - libllvm15.0-15.0.6-1.mga9.i586
> - liblm_sensors5-3.6.0-4.mga9.i586
> - liblzma5-5.4.3-1.mga9.i586
> - libmesaegl1-24.2.8-1.mga9.tainted.i586
> - libmesagl1-24.2.8-1.mga9.tainted.i586
> - libncurses6-6.3-20221203.2.1.mga9.i586
> - libnghttp2_14-1.61.0-1.mga9.i586
> - libnvidia-egl-wayland1-1.1.11-1.mga9.i586
> - libopencl1-2.3.1-2.mga9.i586
> - libopengl0-1.6.0-1.mga9.i586
> - libopenssl3-3.0.15-1.1.mga9.i586
> - libpciaccess0-0.17-1.mga9.i586
> - libpsl5-0.21.2-2.mga9.i586
> - libsasl2_3-2.1.27-7.mga9.i586
> - libssh4-0.10.6-1.mga9.i586
> - libunistring5-1.1-1.mga9.i586
> - libverto1-0.3.2-4.mga9.i586
> - libwayland-client0-1.21.0-1.mga9.i586
> - libwayland-server0-1.21.0-1.mga9.i586
> - libx11-xcb1-1.8.6-1.1.mga9.i586
> - libx11_6-1.8.6-1.1.mga9.i586
> - libxau6-1.0.11-1.mga9.i586
> - libxcb-dri2_0-1.15-2.mga9.i586
> - libxcb-dri3_0-1.15-2.mga9.i586
> - libxcb-glx0-1.15-2.mga9.i586
> - libxcb-present0-1.15-2.mga9.i586
> - libxcb-randr0-1.15-2.mga9.i586
> - libxcb-shm0-1.15-2.mga9.i586
> - libxcb-sync1-1.15-2.mga9.i586
> - libxcb-xfixes0-1.15-2.mga9.i586
> - libxcb1-1.15-2.mga9.i586
> - libxdmcp6-1.1.4-1.mga9.i586
> - libxext6-1.3.5-1.mga9.i586
> - libxfixes3-6.0.0-2.mga9.i586
> - libxshmfence1-1.3.2-1.mga9.i586
> - libxxf86vm1-1.1.5-1.mga9.i586
> - libzlib1-1.2.13-1.2.mga9.i586
> - libzstd1-1.5.5-1.mga9.i586
> - nvidia-current-lib32-550.142-2.mga9.nonfree.x86_64
> 
> 248MB of additional disk space will be used.
> 
> That doesn't make sense to me.

nvidia-current-all is for installing the "full optional" version (like if you were installing the upstream nvidia package), which also includes support for 32bit arch (e.g. for playing 32bit games, consider for instance steam). And for that, the 32bit libraries dependencies are required, that's the long list of libraries you see there.

After playing around with qarepo's wildcard feature, I get the following updated list of 64-bit repos for this bug. Please advise if it is complete and correct.

dkms-nvidia-current-550.142-2.mga9.nonfree.x86_64.rpm
dkms-nvidia-current-open-550.142-2.mga9.nonfree.x86_64.rpm
nvidia-current-all-550.142-2.mga9.nonfree.x86_64.rpm
nvidia-current-cuda-opencl-550.142-2.mga9.nonfree.x86_64.rpm
nvidia-current-devel-550.142-2.mga9.nonfree.x86_64.rpm
nvidia-current-doc-html-550.142-2.mga9.nonfree.x86_64.rpm
nvidia-current-lib32-550.142-2.mga9.nonfree.x86_64.rpm
nvidia-current-utils-550.142-2.mga9.nonfree.x86_64.rpm
x11-driver-video-nvidia-current-550.142-2.mga9.nonfree.x86_64.rpm

mageia-repos-9-3.mga9.x86_64.rpm
mageia-repos-cauldron-9-3.mga9.x86_64.rpm
mageia-repos-keys-9-3.mga9.noarch.rpm
mageia-repos-pkgprefs-9-3.mga9.noarch.rpm

meta-task-9-3.mga9.noarch.rpm

ldetect-lst-0.6.59-1.mga9.x86_64.rpm
ldetect-lst-devel-0.6.59-1.mga9.x86_64.rpm

dkms-nvidia470-470.256.02-2.mga9.nonfree.x86_64.rpm
nvidia470-all-470.256.02-2.mga9.nonfree.x86_64.rpm
nvidia470-cuda-opencl-470.256.02-2.mga9.nonfree.x86_64.rpm
nvidia470-devel-470.256.02-2.mga9.nonfree.x86_64.rpm
nvidia470-doc-html-470.256.02-2.mga9.nonfree.x86_64.rpm
nvidia470-lib32-470.256.02-2.mga9.nonfree.x86_64.rpm
nvidia470-utils-470.256.02-2.mga9.nonfree.x86_64.rpm
x11-driver-video-nvidia470-470.256.02-2.mga9.nonfree.x86_64.rpm

(In reply to Thomas Andrews from comment #15)

> After playing around with qarepo's wildcard feature, I get the following
> updated list of 64-bit repos for this bug. Please advise if it is complete
> and correct.

I don't get the files list anymore from https://madb.mageialinux-online.org/rpmsforqa/33872, files list is empty.

> 
> dkms-nvidia-current-550.142-2.mga9.nonfree.x86_64.rpm
> dkms-nvidia-current-open-550.142-2.mga9.nonfree.x86_64.rpm
> nvidia-current-all-550.142-2.mga9.nonfree.x86_64.rpm
> nvidia-current-cuda-opencl-550.142-2.mga9.nonfree.x86_64.rpm
> nvidia-current-devel-550.142-2.mga9.nonfree.x86_64.rpm
> nvidia-current-doc-html-550.142-2.mga9.nonfree.x86_64.rpm
> nvidia-current-lib32-550.142-2.mga9.nonfree.x86_64.rpm
> nvidia-current-utils-550.142-2.mga9.nonfree.x86_64.rpm
> x11-driver-video-nvidia-current-550.142-2.mga9.nonfree.x86_64.rpm
> 
> mageia-repos-9-3.mga9.x86_64.rpm
> mageia-repos-cauldron-9-3.mga9.x86_64.rpm
> mageia-repos-keys-9-3.mga9.noarch.rpm
> mageia-repos-pkgprefs-9-3.mga9.noarch.rpm
> 
> meta-task-9-3.mga9.noarch.rpm
> 
> ldetect-lst-0.6.59-1.mga9.x86_64.rpm
> ldetect-lst-devel-0.6.59-1.mga9.x86_64.rpm
> 
> dkms-nvidia470-470.256.02-2.mga9.nonfree.x86_64.rpm
> nvidia470-all-470.256.02-2.mga9.nonfree.x86_64.rpm
> nvidia470-cuda-opencl-470.256.02-2.mga9.nonfree.x86_64.rpm
> nvidia470-devel-470.256.02-2.mga9.nonfree.x86_64.rpm
> nvidia470-doc-html-470.256.02-2.mga9.nonfree.x86_64.rpm
> nvidia470-lib32-470.256.02-2.mga9.nonfree.x86_64.rpm
> nvidia470-utils-470.256.02-2.mga9.nonfree.x86_64.rpm
> x11-driver-video-nvidia470-470.256.02-2.mga9.nonfree.x86_64.rpm

Yes, list seems correct, of course not all the packages from the list above can be installed at one shot all together at the same time.

@Giuseppe , before I upload a list, I find nvidia cuda packages, they should be included in this bug or you will open a new one for them?

(In reply to Giuseppe Ghibò from comment #16)

> 
> Yes, list seems correct, of course not all the packages from the list above
> can be installed at one shot all together at the same time.

I know, but I needed them in a local repo to test that the issue in comment 9 has been resolved. So my first test went like this:

I was attempting to simulate a relatively inexperienced user who had just purchased/installed a Quadro k620 card. To set this up, I removed all nvidia packages from the system, and booted using nouveau (which works OK for "ordinary" usage). Then I went after updates, ldetect-lst being the only one waiting, because I had already updated meta-task and mageia-repos for comment 9.

Next was to run MCC and install the recommended driver - nvidia-current. There were no issues there, and none after reboots into both desktop and server kernels. This first test is to check just the packages installed by XFdrake, so no cuda/openCL packages.

Next was an attempt by this fictional inexperienced user to try out nvidia470, again as installed through MCC. This time there was no conflict, the driver was installed, and seemed to work with both kernels. Switching back to nvidia-current also went without issues.

So, the issue from comment 9 appears to be resolved. But, I have not tested the cuda packages yet. I'm something of a neophyte in this area. Could someone suggest a relatively simple way that I could test that the various cuda/openCL packages are making a difference?

(In reply to Thomas Andrews from comment #18)

> (In reply to Giuseppe Ghibò from comment #16)
> 
> > 
> > Yes, list seems correct, of course not all the packages from the list above
> > can be installed at one shot all together at the same time.
> 
> I know, but I needed them in a local repo to test that the issue in comment
> 9 has been resolved. So my first test went like this:
> 
> I was attempting to simulate a relatively inexperienced user who had just
> purchased/installed a Quadro k620 card. To set this up, I removed all nvidia
> packages from the system, and booted using nouveau (which works OK for
> "ordinary" usage). Then I went after updates, ldetect-lst being the only one
> waiting, because I had already updated meta-task and mageia-repos for
> comment 9.
> 
> Next was to run MCC and install the recommended driver - nvidia-current.
> There were no issues there, and none after reboots into both desktop and
> server kernels. This first test is to check just the packages installed by
> XFdrake, so no cuda/openCL packages.
> 
> Next was an attempt by this fictional inexperienced user to try out
> nvidia470, again as installed through MCC. This time there was no conflict,
> the driver was installed, and seemed to work with both kernels. Switching
> back to nvidia-current also went without issues.

ok, so switching back and forth seems ok, which is a bit more robust than before. IMHO would work also work switching back and forth between all the 5 drivers (including -newfeaure, modesetting and nouveau) at least in this card, but this is probably beyond this bug.

> 
> So, the issue from comment 9 appears to be resolved. But, I have not tested
> the cuda packages yet. I'm something of a neophyte in this area. Could
> someone suggest a relatively simple way that I could test that the various
> cuda/openCL packages are making a difference?

for cuda a simple test is cuda-z (works only in nvidia-current or nvidia-newfeature, not nvidia470) which would tell the speed of the card if cuda is correctly installed. Another test is contained in the bunch of tests into nvidia-cuda-toolkit-samples-bins package.

A further another real test is a render in blender using cycles with specifying CUDA in blender's system renderer, but that requires more knowledge (at least basic of blender) and then pressing F12 for rendering.

(In reply to katnatek from comment #17)

> @Giuseppe , before I upload a list, I find nvidia cuda packages, they should
> be included in this bug or you will open a new one for them?

nvidia-cuda packages which one you mean exactly? You mean the nvidia-cuda-toolkit or another one?

For nvidia-cuda-toolkit, there is version 12.2 in updates testing, but was old, so worthwhile to skip to the latest cuda for series 550.xx, which should be around 12.3 or 12.4, and then pass to latest absolute (12.6u3) ,which however none of them is actually already ready.

(In reply to Giuseppe Ghibò from comment #19)

> for cuda a simple test is cuda-z (works only in nvidia-current or
> nvidia-newfeature, not nvidia470) which would tell the speed of the card if

for cuda-z for nvidia470 a trick is to download the upstream cuda-z from sourceforge, (however it's no longer maintained), that was a self-extracting .run archive. A bit messy the GUI with the fonts, but should start, as it was compiled with an older (7.x or so) CUDA.

Created attachment 14823 [details]
Current file list

I hope not miss something

Attachment 14817 is obsolete: 0 => 1

Installed nvidia-current-cuda-opencl, then cuda-z and the toolkit from the nonfree repos. I'll post the exported cuda-z text file, but to my untrained eye it looks OK.

I also installed Darktable. When I ran it for the first time, it showed messages near the bottom of the display that indicated it was being configured for cuda, but most of the messages were too fast to read. It finished without incident.

My Darktable skills are pretty much limited to loading a set of images. I did that, and ran a slide show, and it was all OK.

That's about as far as I can take it.

Created attachment 14824 [details]
cuda-z report from Quadro K620 and nvidia-current

(In reply to Thomas Andrews from comment #24)
> Created attachment 14824 [details]
> cuda-z report from Quadro K620 and nvidia-current

It seems everything is OK.

For your data output, it tell for instance the bandwidth is of 12 GB/s device to device and 6 GB/s device to host. With FP32 operations at 63 GFLOPS, and FP64 at 13 GFLOPS.

By the same math, RTX 4090 should be for instance around 82 TFLOPS FP32 and 1.2 TFLOPS FP64. H100 which are designed for AI or HPC should be around 51 TFLOPS FP32 and 25 TFLOPS FP64.

Created attachment 14825 [details]
470 cuda-z report

cuda-z report for nvidia470, using downloaded cuda-z. This one loooks OK, too.

When run, Darktable re-configured itself foe the nvidia470 cuda, as well.

Now what we need is someone with the right hardware to test dkms-nvidia-current-open...

I added a point under
https://wiki.mageia.org/en/Mageia_10_Release_Notes#Proprietary_NVIDIA_driver

I think the text there should be elaborated, later.

Keywords: FOR_RELEASENOTES10 => IN_RELEASENOTES10

(In reply to Morgan Leijström from comment #29)
> I added a point under
> https://wiki.mageia.org/en/Mageia_10_Release_Notes#Proprietary_NVIDIA_driver
> 
> I think the text there should be elaborated, later.

BTW, in the doc there is also some comment about nvidia390, it was added a procedure to use it, outside mageia official drivers, here:
https://bugs.mageia.org/show_bug.cgi?id=33549#c13, though there wasn't any feedback.

Noted there :-)

I noticed there is an unexpected conflicts, when nvidia-current-all is installed and install dkms-nvidia-current-open. I've fixed already in cauldron in 550.142-4, so worthwhile to use the fix also for mga9, which there will be 550.142-3.

Created attachment 14829 [details]
updated files list

Attachment 14823 is obsolete: 0 => 1

I've updated the files list of katnatek to the newer binaries.

For the usage of the -open kernel modules variant:

- configure the nvidia-current driver as usual from XFrake (it will install dkms-nvidia-current), and reboot.

- install dkms-nvidia-current-open (this would uninstall dkms-nvidia-current-open), then reboot.

I ran another battery of tests on the newer binaries, with the same results. I also have updated nvidia-current on my production install of this hardware. I will use it for a few days, and if any issues show up, I'll be sure to report them.

MGA9-64, GNOME, AMD Ryzen 5600, Nvidia 1050


The following 7 packages are going to be installed:

- dkms-nvidia-current-550.142-3.mga9.nonfree.x86_64
- nvidia-current-all-550.142-3.mga9.nonfree.x86_64
- nvidia-current-cuda-opencl-550.142-3.mga9.nonfree.x86_64
- nvidia-current-doc-html-550.142-3.mga9.nonfree.x86_64
- nvidia-current-lib32-550.142-3.mga9.nonfree.x86_64
- nvidia-current-utils-550.142-3.mga9.nonfree.x86_64
- x11-driver-video-nvidia-current-550.142-3.mga9.nonfree.x86_64

- ldetect-lst-0.6.59-1.mga9.x86_64
- mageia-repos-9-3.mga9.x86_64
- mageia-repos-keys-9-3.mga9.noarch
- mageia-repos-pkgprefs-9-3.mga9.noarch



---rebooted

- Nvidia working

$ nvidia-smi
Fri Jan  3 13:08:16 2025       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.142                Driver Version: 550.142        CUDA Version: 12.4     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce GTX 1050        Off |   00000000:05:00.0  On |                  N/A |
| 45%   26C    P8             N/A /   75W |     101MiB /   2048MiB |      1%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+





- system behaving as expected.

Advisory updated to latest list

I would rather have specific tests of the new "open" driver, but it doesn't look like that's going to happen. The three proprietary drivers seem to be working OK.

Validating.

CC: (none) => sysadmin-bugs
Keywords: (none) => validated_update

An update for this issue has been pushed to the Mageia Updates repository.

https://advisories.mageia.org/MGAA-2025-0001.html

Status: NEW => RESOLVED
Resolution: (none) => FIXED

Change keyword to flag status

Flags: (none) => in_release_notes10+
Keywords: IN_RELEASENOTES10 => (none)