Page MenuHomeSolus

NVidia 390 driver is glitchy
Closed, WontfixPublic

Description

After recent update I noticed:

  • My cursor is lagging when I move over youtube videos (a copy of the cursor remains back and is visible about one second).
  • Google maps satellite 3D mode freezes a few seconds on load, then shows black areas. At the end everything is loaded and the rotation is smooth.
  • My Chrome screenshot extension doesn't show the captured screenshot (it uses html5 canvas to display the screenshot).

In general when there is some video/3D somewhere I notice the cursor glitching/freezing for a few milliseconds.


I removed all nvidia-glx-driver* packages and installed the nvidia-340 or the nvidia-304 drivers but both make my laptop not starting up (it shows black screen with blinking underscore).
Then via Ctrl+Alt+F1 I installed nvidia-glx-driver* and it started again but the glitches are still there.


So the nvidia-340 and nvidia-304 drivers are not usable.
I would like to be able to install the nvidia-38x driver, until the problem with 390 will be fixed.


  • It happens both with linux-current and linux-lts
  • My video card model is 960M

Event Timeline

arteniioleg updated the task description. (Show Details)
arteniioleg updated the task description. (Show Details)Feb 3 2018, 4:30 PM
$ steam
Running Steam on solus 3 64-bit
STEAM_RUNTIME is disabled by the user
Installing breakpad exception handler for appid(steam)/version(1513371133)
libGL error: No matching fbConfigs or visuals found
libGL error: failed to load driver: swrast
Installing breakpad exception handler for appid(steam)/version(1513371133)
Installing breakpad exception handler for appid(steam)/version(1513371133)
Gtk-Message: Failed to load module "canberra-gtk-module"
Installing breakpad exception handler for appid(steam)/version(1513371133)
libGL error: No matching fbConfigs or visuals found
libGL error: failed to load driver: swrast
glXChooseVisual failedMain.cpp (322) : Assertion Failed: Fatal Error: glXChooseVisual failed
Assert( Assertion Failed: Fatal Error: glXChooseVisual failed ):Main.cpp:322

Installing breakpad exception handler for appid(steam)/version(1513371133)
assert_20180203183334_6.dmp[2332]: Uploading dump (out-of-process)
/tmp/dumps/assert_20180203183334_6.dmp
assert_20180203183334_6.dmp[2332]: Finished uploading minidump (out-of-process): success = yes
assert_20180203183334_6.dmp[2332]: response: CrashID=bp-7de1c994-577c-4a20-bfaa-eeb552180203
assert_20180203183334_6.dmp[2332]: file ''/tmp/dumps/assert_20180203183334_6.dmp'', upload yes: ''CrashID=bp-7de1c994-577c-4a20-bfaa-eeb552180203''

arteniioleg updated the task description. (Show Details)Feb 3 2018, 4:47 PM

So the nvidia-340 and nvidia-304 drivers are not usable.

These drivers are for older generation GeForce GPU and they are perfectly usable... with the correct hardware ?

arteniioleg added a comment.EditedFeb 3 2018, 6:14 PM

I managed to install 384 from the nvidia site and it works smooth without any mouse, google maps, youtube lagging (chrome screenshot extension and steam works too).

One downside is that it doesn't show up in sudo eopkg list-installed | grep nvidia I don't know how I will uninstall it.

mati865 added a subscriber: mati865.Feb 3 2018, 8:25 PM

Installing Nvidia drivers manually does huge mess in the system.
To uninstall it run installer (the same version as one used to install) with --uninstall argument and reinstall mesalib, libglvnd, xorg-server via eopkg. Then install drivers with Linux Driver Manager.

You haven't said which processor you have got but I'm guessing it's Intel with HD Graphics.
I don't know how exactly Solus handles Optimus systems so let's wait for someone who actually knows.
I can only guess there was some kind of conflict between mesa and nvidia but it would require more details like list of loaded libs by process to be sure.

arteniioleg added a comment.EditedFeb 5 2018, 6:38 PM

@mati865 Thanks, I uninstalled it with --uninstall.

A few minutes ago I installed the latest updates, I saw the nvidia 390 drivers were in the list. But the glitch still persists (also after reinstalling mesalib, libglvnd, xorg-server).

Could you provide output of inxi -G?

Sure, anything. (The main reason I am still on Solus it's because of the best 3D performance compared to any other distro I've used, and now this is fading).

Graphics:  Card-1: Intel HD Graphics 530
           Card-2: NVIDIA GM107M [GeForce GTX 960M]
           Display Server: x11 (X.Org 1.19.6 ) drivers: modesetting,nvidia
           Resolution: 3840x2160@60.00hz
           OpenGL: renderer: GeForce GTX 960M/PCIe/SSE2
           version: 4.5.0 NVIDIA 390.25

P.S. More info is in the above dump file (at the end of the file there is text information)

I don't know how to open Steam crash dumps (probably one would need their specialized software).

Are you running fully upgraded system (sudo eopkg upgrade)?
I cannot reproduce it on the desktop with latest packages and Nvidia 390 branch.

Yes I tried eopkg upgrade, it said nothing to upgrade.

My machine is Dell XPS 9550.

In the past 2 days I searched for a distro to migrate to. I installed ElementaryOS, Ubuntu 17.10, Linux Mint but none look so slick like Budgie and 3D performace is bad (low fps).
So now I am back to Solus (fresh reinstall) and this problem persists.

Strange that it has a negative and I positive side:

  • Negative: When there is some video/3D somewhere, the cursor/system is lagging until that video/3D is fully initialized.
  • Positive: Once that video/3D is initialized, it's performance is the best I ever experienced (high FPS, I can notice this clearly in google maps 3D - the globe rotation is very smooth)

There is something weird going on with Nvidia dGPU recently: https://dev.solus-project.com/T5701

@ikey any thoughts?

I just installed latest updates and the problem persists. Here is a video

I just noticed that the video doesn't show what I see on the screen: 2 mouse pointers, one is still/blocked and another is currently my mouse moving. When that happens the video recording freezes completely.

arteniioleg added a comment.EditedFeb 18 2018, 7:20 PM

Some html5 applications that use <canvas> are glitching because of this bug. For example canva.com doesn't display some image elements and the browser screenshot extension shows a black screen instead of the captured page https://streamable.com/3aqa7

Also in nVidia Settings there is no option to switch to Intel video card.

I wasn't able to do my work in canva.com so I uninstalled nVidia drivers and restarted. But it performed way worse: web page elements were disappearing, slow rendering piece by piece and appearing after I scroll or browser window resize.

Latest Friday updates are installed.

I tried to install the latest driver but after restart the desktop was black, no status bar and windows didn't have a menu bar.

Please push 390.42 in latest software updates, maybe it will help fix this issue.

packetherder added a subscriber: packetherder.EditedMar 20 2018, 3:26 AM

Here's the link on Nvidia's forums talking about similar issues with 390.25 and chrome. Doesn't seem like 390.42 really helps either. I've been getting by with just disabling hardware acceleration, but that's less than ideal.

quote: None of these issues occur when using version 387.34.

Why did Solus remove 387 from repo? I can't find it with sudo eopkg list-available | grep nvidia.
Just reverting to 387 would be a solution (but I can't do that, as already described in previous comments)

Maybe this is the fix

Further improved the fix for occasional flicker when using the X driver's composition pipeline. This was mostly fixed in 390.42, but now the fix should be more complete.
http://www.nvidia.com/download/driverResults.aspx/132530/

Can you push this version to repo please?

Also people in comments are saying that the latest 390.x version has visible improvements https://devtalk.nvidia.com/default/topic/1029484/linux/-various-all-distros-numerous-performance-amp-rendering-issues-on-390-25/post/5249705/#5249705

cc @packetherder

Thank you for the update but I got black screen with a blinking underscore after reboot.

I uninstalled all nvidia packages via Ctrl+Alt+F1

Again, allowing us to go back to nvidia-38x will be better (it worked well for me)

The above comment is what happened at my office PC. That PC is not the one I was talking about in all the above comments. My work PC worked just fine with 390 driver until today's update, when I got black screen with blinking underscode (second reboot Ctrl+Alt+Delete didn't help).

So now I am on my home laptop, the one this issue is about (Dell XPS 15 9550). I just installed the latest updates (nvidia 390.48). Fortunately I didn't get black screen after reboot, everything works the same, including the mouse freezing glitch (this bug persists).

On my office PC I installed 340 but I also got black screen with blinking underscore (then I uninstalled again all nvidia packages via Ctrl+Alt+F1).

So most nvidia drivers are not usable.

So most nvidia drivers are not usable.

I'm running on latest nvidia drivers (390 now) and had no problems at all. I'm out of ideas.

What commands should I type to provide you useful debug info?

Somebody using laptop with working nvidia drivers could compare (or post) their Xorg.log, journal and dmesg with nvidia-bug-report.log.gz posted earlier in this thread. I cannot help any further.

What kernel are you running where you are getting the black screen?

uname -a

Linux xu 4.15.15-63.current #1 SMP PREEMPT Sat Apr 7 08:32:34 UTC 2018 x86_64 GNU/Linux

Are you fully up to date? Seems the kernel is a sync behind. If you aren't running the latest kernel/drivers when the drivers are first installed then a black screen will happen regardless which version you are using.

If your repo is up to date and your system up to date, you should be booting into the 4.15.17-64 kernel

I just installed nvidia-390 and got black screen. Then I went to Ctrl+Alt+F1 and generated all logs specified in https://dev.solus-project.com/T5727#112573

arteniioleg added a comment.EditedApr 18 2018, 7:39 AM

Sure, I am always up-to-date. I open Software Center > Updates > click Check for updates

Also from terminal

xu@xu ~ $ uname -a
Linux xu 4.15.15-63.current #1 SMP PREEMPT Sat Apr 7 08:32:34 UTC 2018 x86_64 GNU/Linux
xu@xu ~ $ sudo eopkg update-repo 
Password: 
Updating repository: Solus
eopkg-index.xml.xz.sha1sum     (40.0  B)100%    709.26 KB/s [00:00:00] [complete]
Solus repository information is up-to-date.
xu@xu ~ $ sudo eopkg upgrade 
Updating repositories
Updating repository: Solus
eopkg-index.xml.xz.sha1sum     (40.0  B)100%    507.24 KB/s [00:00:00] [complete]
Solus repository information is up-to-date.
No packages to upgrade.

Your latest logs contain clue:
dmesg:

[    2.746003] NVRM: API mismatch: the client has the version 390.48, but
               NVRM: this kernel module has the version 390.25.  Please
               NVRM: make sure that this kernel module and all NVIDIA driver
               NVRM: components have the same version.

Try eopkg upgrade.

I just did that in the previous comment.

xu@xu ~ $ sudo eopkg update-repo --force
Updating repository: Solus
eopkg-index.xml.xz.sha1sum     (40.0  B)100%    952.56 KB/s [00:00:00] [complete]
Solus repository information is up-to-date.
Updating database at any rate as requested
eopkg-index.xml.xz             (1.8 MB)100%     56.02 MB/s [00:00:00] [complete]
No signature found for https://packages.solus-project.com/shannon/eopkg-index.xml.xz
Package database updated.
xu@xu ~ $ sudo eopkg upgrade
Updating repositories
Updating repository: Solus
eopkg-index.xml.xz.sha1sum     (40.0  B)100%    930.91 KB/s [00:00:00] [complete]
Solus repository information is up-to-date.
No packages to upgrade.
xu@xu ~ $ uname -a
Linux xu 4.15.15-63.current #1 SMP PREEMPT Sat Apr 7 08:32:34 UTC 2018 x86_64 GNU/Linux

The new kernel is installed in /usr but not in the bootloader. This will tell you what's going on

sudo CBM_DEBUG=1 clr-boot-manager update

If you are using a different bootloader/dual boot, that needs to be updated to reflect the latest kernel.

xu@xu ~ $ sudo CBM_DEBUG=1 clr-boot-manager update
Password: 
[INFO] cbm (src/bootman/bootman.c:L437): Current running kernel: 4.15.15-63.current
[INFO] cbm (src/bootman/sysconfig.c:L98): Discovered UEFI ESP: /dev/disk/by-partuuid/c8026340-bb49-45ec-ba62-977a050f2cb5
[INFO] cbm (src/bootman/sysconfig.c:L123): Fully resolved boot device: /dev/sda5
[DEBUG] cbm (src/bootman/bootman.c:L124): UEFI boot now selected (systemd)
[INFO] cbm (src/bootman/update.c:L75): Checking for mounted boot dir
[INFO] cbm (src/bootman/update.c:L108): Mounting boot device /dev/sda5 at /boot
[SUCCESS] cbm (src/bootman/update.c:L116): /dev/sda5 successfully mounted at /boot
[DEBUG] cbm (src/bootman/update.c:L245): Now beginning update_native
[DEBUG] cbm (src/bootman/update.c:L254): update_native: 3 available kernels
[DEBUG] cbm (src/bootman/update.c:L274): update_native: Running kernel is (current) ///usr/lib/kernel/com.solus-project.current.4.15.15-63
[SUCCESS] cbm (src/bootman/update.c:L289): update_native: Bootloader updated
[SUCCESS] cbm (src/bootman/update.c:L298): update_native: Repaired running kernel ///usr/lib/kernel/com.solus-project.current.4.15.15-63
[DEBUG] cbm (src/bootman/update.c:L307): update_native: Checking kernels for type current
[INFO] cbm (src/bootman/update.c:L322): update_native: Default kernel for type current is ///usr/lib/kernel/com.solus-project.current.4.15.17-64
[FATAL] cbm (src/bootman/kernel.c:L654): Failed to install initrd /boot/EFI/com.solus-project/initrd-com.solus-project.current.4.15.17-64: No space left on device
[FATAL] cbm (src/bootman/update.c:L329): Failed to install default-current kernel: ///usr/lib/kernel/com.solus-project.current.4.15.17-64
[INFO] cbm (src/bootman/update.c:L134): Attempting umount of /boot
[SUCCESS] cbm (src/bootman/update.c:L138): Unmounted boot directory
sudo fdisk -l

...

Device         Start       End   Sectors   Size Type
/dev/sda1       2048    114687    112640    55M BIOS boot
/dev/sda2     311296  39178239  38866944  18.5G Linux filesystem
/dev/sda3   39178240 484491263 445313024 212.4G Linux filesystem
/dev/sda4  484491264 488396799   3905536   1.9G Linux swap
/dev/sda5     114688    311295    196608    96M EFI System

In this case, with office PC, it's my fault that I missed this phrase from docs:

To create an EFI System Partition, open up GParted and create a FAT32 partition that is 512MB in size.
https://solus-project.com/articles/installation/disks/en/

Case for office PC is closed. Thank you!
I will reinstall Solus and recreated the first boot partition with required size.

I will check my home PC if it has the same problem.

At my home laptop everything is fine with the boot partition

o@a ~ $ uname -a
Linux a 4.15.17-64.current #1 SMP PREEMPT Thu Apr 12 18:33:03 UTC 2018 x86_64 GNU/Linux
o@a ~ $ sudo CBM_DEBUG=1 clr-boot-manager update
Password: 
[INFO] cbm (src/bootman/bootman.c:L437): Current running kernel: 4.15.17-64.current
[INFO] cbm (src/bootman/sysconfig.c:L98): Discovered UEFI ESP: /dev/disk/by-partuuid/d69e550f-c861-4bd6-a082-8242f2b6f486
[INFO] cbm (src/bootman/sysconfig.c:L123): Fully resolved boot device: /dev/sda1
[DEBUG] cbm (src/bootman/bootman.c:L124): UEFI boot now selected (systemd)
[INFO] cbm (src/bootman/update.c:L75): Checking for mounted boot dir
[INFO] cbm (src/bootman/update.c:L108): Mounting boot device /dev/sda1 at /boot
[SUCCESS] cbm (src/bootman/update.c:L116): /dev/sda1 successfully mounted at /boot
[DEBUG] cbm (src/bootman/update.c:L245): Now beginning update_native
[DEBUG] cbm (src/bootman/update.c:L254): update_native: 2 available kernels
[DEBUG] cbm (src/bootman/update.c:L274): update_native: Running kernel is (current) ///usr/lib/kernel/com.solus-project.current.4.15.17-64
[SUCCESS] cbm (src/bootman/update.c:L289): update_native: Bootloader updated
[SUCCESS] cbm (src/bootman/update.c:L298): update_native: Repaired running kernel ///usr/lib/kernel/com.solus-project.current.4.15.17-64
[DEBUG] cbm (src/bootman/update.c:L307): update_native: Checking kernels for type current
[INFO] cbm (src/bootman/update.c:L322): update_native: Default kernel for type current is ///usr/lib/kernel/com.solus-project.current.4.15.17-64
[SUCCESS] cbm (src/bootman/update.c:L334): update_native: Installed tip for current: ///usr/lib/kernel/com.solus-project.current.4.15.17-64
[SUCCESS] cbm (src/bootman/update.c:L348): update_native: Installed last_good kernel (current) (///usr/lib/kernel/com.solus-project.current.4.15.17-64)
[DEBUG] cbm (src/bootman/update.c:L359): update_native: Analyzing for type current: ///usr/lib/kernel/com.solus-project.current.4.15.17-64
[DEBUG] cbm (src/bootman/update.c:L362): update_native: Skipping running kernel
[DEBUG] cbm (src/bootman/update.c:L359): update_native: Analyzing for type current: ///usr/lib/kernel/com.solus-project.current.4.12.7-11
[INFO] cbm (src/bootman/update.c:L365): update_native: not-running: ///usr/lib/kernel/com.solus-project.current.4.12.7-11
[INFO] cbm (src/bootman/update.c:L375): update_native: not-default-current: ///usr/lib/kernel/com.solus-project.current.4.12.7-11
[INFO] cbm (src/bootman/update.c:L381): update_native: not-last-booted: ///usr/lib/kernel/com.solus-project.current.4.12.7-11
[INFO] cbm (src/bootman/update.c:L392): update_native: Proposed for deletion from current: ///usr/lib/kernel/com.solus-project.current.4.12.7-11
[SUCCESS] cbm (src/bootman/update.c:L416): update_native: Default kernel for current is ///usr/lib/kernel/com.solus-project.current.4.15.17-64
[INFO] cbm (src/bootman/update.c:L432): update_native: Garbage collecting current: ///usr/lib/kernel/com.solus-project.current.4.12.7-11
[INFO] cbm (src/bootman/update.c:L134): Attempting umount of /boot
[SUCCESS] cbm (src/bootman/update.c:L138): Unmounted boot directory



Here is a screencast of how my mouse (everything) freezes when some 3D/Video content is initialized/loaded

Doesn't look good:

Apr 16 11:06:30 a nvidia-persistenced[599]: Failed to query NVIDIA devices. Please ensure that the NVIDIA device files (/dev/nvidia*) exist, and that user 143 has read and write permissions for those files.
...
Apr 16 11:07:38 a kernel: nvidia-modeset: Allocated GPU:0 (GPU-370ccae8-0775-8452-4bee-a9b6b7f874e9) @ PCI:0000:01:00.0
Apr 16 11:07:38 a kernel: nvidia-modeset: Freed GPU:0 (GPU-370ccae8-0775-8452-4bee-a9b6b7f874e9) @ PCI:0000:01:00.0

Could you check if you have anything like nvidia inside /dev/?

o@a ~ $ ls -l /dev/ | grep nvidia
crw-rw-rw-  1 root root    195,   0 Apr 18 21:45 nvidia0
crw-rw-rw-  1 root root    195, 255 Apr 18 21:45 nvidiactl
crw-rw-rw-  1 root root    195, 254 Apr 18 21:45 nvidia-modeset
crw-rw-rw-  1 root root    241,   0 Apr 18 21:45 nvidia-uvm

... and that user 143 has read and write permissions for those files.

So I should change the owner of those files from root to UID 143?

There should be line nvidia-persistenced:x:143: in /etc/group.
If it's not there then something went terribly bad and other issues can be harder to fix.

It's there (the last line)

o@a ~ $ sudo cat /etc/group
Password: 
root:x:0:
bin:x:1:
sys:x:2:
kmem:x:3:
tape:x:4:
tty:x:5:
daemon:x:6:
floppy:x:7:
disk:x:8:
lp:x:9:
dialout:x:10:o
audio:x:11:o
video:x:12:o
utmp:x:13:
usb:x:14:
cdrom:x:15:o
mail:x:34:
nogroup:x:99:
wheel:x:1001:
lock:x:54:
messagebus:x:18:
pulse:x:58:
sudo:x:28:o
fuse:x:104:o
adm:x:997:
saned:x:991:
systemd-timesync:x:992:
pulse-access:x:59:
systemd-network:x:990:
users:x:100:o
colord:x:124:
systemd-journal:x:996:
systemd-resolve:x:995:
systemd-bus-proxy:x:994:
nobody:x:65534:
nm-openvpn:x:993:
input:x:999:
polkitd:x:27:
avahi:x:84:
lpadmin:x:19:o
lightdm:x:620:
o:x:1000:
scanner:x:70:o
plugdev:x:90:o
systemd-coredump:x:988:
kvm:x:989:
docker:x:145:o
nm-openconnect:x:998:
avahi-autoipd:x:987:
sambaguest:x:65533:
nvidia-persistenced:x:143:

So it looks like Nvidia forums is the last hope.

Hi guys,

I found this page while looking for solution to solve Chrome screenshot stop working after installing Nvidia 390.48 driver (even 396.18 beta not work).
Since I use Ubuntu 17.10, there's nothing I can help on installing driver on Solus, but I can offer some suggestion on Chrome settings.

My machine is Dell Inspiron 5557
here is the output of inxi -G

Graphics:  Card-1: Intel HD Graphics 520
           Card-2: NVIDIA GM108M [GeForce 930M]
           Display Server: x11 (X.Org 1.19.5 ) drivers: modesetting,nvidia Resolution: 1920x1080@60.00hz
           OpenGL: renderer: GeForce 930M/PCIe/SSE2 version: 4.5.0 NVIDIA 396.18

I followed these steps to install Nvidia drivers, no black login screen with lightdm

  1. add graphics-drivers/ppa repository
  2. install nvidia-prime, ubuntu-drivers-common, mesa-utils, and driver nvidia-3xx
  3. sudo prime-select intel then reboot
  4. sudo prime-select nvidia then reboot, now you have enabled Nvidia driver

Everything works fine on Nvidia 387.34 driver, no problem with Chrome.

After upgrading to Nvidia 390.48 driver, Chrome screenshot always give me a black image.

Finally I found some workaround on Gliffy Troubleshooting Disappearing Shapes And Lines to solve this problem.

So here is my Chrome execute parameters (enable all gpu feature except 2D canvas hardware acceleration)
/usr/bin/google-chrome --ignore-gpu-blacklist --enable-gpu-rasterization --enable-native-gpu-memory-buffers --enable-features="CheckerImaging" --disable-accelerated-2d-canvas

chrome://gpu result as following (screenshot works as before, Google Map 3D runs smoothly even on Nvidia 396.18, Chrome 66.0.3359.117)

Cheers,
Chien

I want a stable driver, not hacks for every program I use.

@cwchien Everything works fine on Nvidia 387.34 driver, no problem with Chrome.

That ppa is not compatible with Solus so I am stuck with only drivers from Solus repo.
Unfortunately there was quiet when I asked them to downgrade to 387 (it worked fine for me too), or to add it as an install option in

sudo eopkg list-available | grep nvidia

@arteniioleg I created a Solus VM on VirtualBox and locally built the nvidia-387-glx-driver (yes, I branched a new package to prevent being forced upgrade).

The package is based on nvidia-glx-driver (390.48-233) and applied patches from "patch for 387.34 and Linux 4.15.0-rc1" on Nvidia forum.

I can install the local packages successfully on Solus stable (kernel 4.15.17), but not sure it works on real Nvidia card since I used a VM.

So if you don't mind, the eopkgs can be download from My Google Drive, please give them a try.

I look forward to some good news from you. :)

Cheers,
Chien

@cwchien It flies! Thank you!


First I had to uninstall all nvidia-glx-* packages because I got conflicting files on 387 packages install.


Just to confirm that I have it installed:

o@a ~ $ sudo eopkg list-installed | grep nvidia
nvidia-304-glx-driver-modaliases            - Files to help with DoFlicky driver detection
nvidia-340-glx-driver-modaliases            - Files to help with DoFlicky driver detection
nvidia-387-glx-driver                       - NVIDIA 387xx Binary Driver (LTS Kernel)
nvidia-387-glx-driver-32bit                 - 32-bit libraries for nvidia-387-glx-driver
nvidia-387-glx-driver-common                - Shared assets for the NVIDIA 387xx GLX Driver
nvidia-387-glx-driver-current               - NVIDIA 387xx Binary Driver (Current Kernel)
nvidia-387-glx-driver-modaliases            - Files to help with DoFlicky driver detection

@arteniioleg Glad to see Nvidia GTX 960M flies again! :)

Conflict with nvidia-glx-driver-common was made on purpose to prevent having two version of drivers at the same time.

Since I'm an Ubuntu user not Solus one, I had better offer the files so that you can build on your own for new kernel.

Here is the packaging files I used, after tar zxvf there is a Packaging folder and just put it in your Home directory.
It contains:

├── common/                                    (official packaging scripts)
├── Makefile -> common/Makefile.toplevel       (symbolic link shared among packages)
├── Makefile.common -> common/Makefile.common  (as above)
├── Makefile.iso -> common/Makefile.iso        (as above)
└── nvidia-387-glx-driver/                     (Nvidia 387xx driver packaging folder)

Basicly, I followed the Solus Packaging Guide and some trail-and-error,

  1. create ~/.solus/packager file and setting your name and email
  2. sudo eopkg it -c system.devel
  3. sudo eopkg it solbuild (we only build package for stable)
  4. sudo solbuild init at the first time, sudo solbuild update for subsequent updates
  5. I have done the folders so that you can just skip the common section
  6. sudo solbuild build package.yml -p main-x86_64 in the nvidia-387-glx-driver/ folder (without -p main-x86_64 will buid for unstable)
  7. if everything goes well, you will see five eopkg files in the nvidia-387-glx-driver/ folder

BTW, if you want to unlock full GPU power on Chrome, these parameter still applied (I took off the --disable-accelerated-2d-canvas since it's not needed for 387.34)
/usr/bin/google-chrome --ignore-gpu-blacklist --enable-gpu-rasterization --enable-native-gpu-memory-buffers --enable-features="CheckerImaging"

Cheers,
Chien

I have successfully built and installed the driver following your steps. Thank you!

nVidia drivers are closed source and Solus developers can't fix them, I am happy with 387, so I'm closing this issue.

arteniioleg closed this task as Wontfix.Apr 30 2018, 4:54 PM
cwchien added a comment.EditedMay 3 2018, 3:26 PM

@arteniioleg Hi, I tried the latest Nvidia 396.24 driver on my Ubuntu, and it seems fixed the screenshot issue and performance is also well, too.

I think maybe you could give 396.24 a try. :)

I installed 396.24 from Solus repo and I got black screen with blinking underscore after restart.
All updates are installed.

I recompiled 387 according to these steps https://dev.solus-project.com/T5727#113009
and I also got black screen with blinking underscore ?

Well at this point you have multiple branches of drivers and custom installs - so its going to be hard to separate the true issues here

I installed 396.24 .run from nvidia.com - after reboot all windows were tiny (no scaling), no 3D acceleration, no header bars on all windows.
After uninstalling it .run --uninstall and installed 396.24 from Software Center. Now it works but has the same mouse freezing glitch as 390.
I will stick to it, the mouse glitch is something minor now compared to the hassle of compiling and fixing the system.

We expressly do not support NVIDIAs .run method for manual installation as it will always break something. I'd consult eopkg check at this point in time, because glvnd egl-wayland etc are all going to have issues.
And there is no real way of knowing what the manual install left behind.

@arteniioleg

In the beginning, I upgraded my Ubuntu to 18.04 (with Kernel 4.15) and found Nvidia 396.24 works very well.

I didn't notice that Ubuntu 18.04 switched back to Xorg from Wayland, and some different configuration also might cause problems.

Sorry to mess up your system. :(

arteniioleg added a comment.EditedMay 27 2018, 5:06 AM

Another observation:

  1. I installed latest updatest 390.59
  2. After restart I got black screen with blinking underscore
  3. I uninstalled all nvidia-* packages via Ctrl+Alt+F1
  4. After restart it works but no 3D acceleration
  5. I installed graphics drivers via DoFlicky
  6. After restart everything works well (the mouse freezing persists though)
  7. After restart I get mouse artifacts like in the previous comment video
  8. After restart everything works fine

I was using Solus because of neat design and 3D performance.
Now I got all of these with Ubuntu 18.04 + Budgie Desktop (plus stability and .deb packages).
On my laptop 3D performace is the same, but most importantly there is no this freezing glitch.
On my desktop PC the 3D performance is better, I play a lot Dirt Rally and on Solus it worked smoothly only on Very Low graphics preset, now on Ubuntu it works ok on Medium preset.
The nvidia driver version is 390, installed from default repository (Settings > Additional Drivers).
P.S. On budgie-desktop package install I chose lightdm.

DataDrake changed the edit policy from "All Users" to "Administrators".Jul 13 2018, 11:12 AM