NVIDIA Optimus isn't working for 470xx

ualkyire · 2022-04-22 15:02:44

Ok, here's the latest Xorg.0.log.

V1del · 2022-04-22 15:14:24

You are loading 390xx there, prime-run will not work with that.

ualkyire · 2022-04-22 15:15:56

Ok, I'll try switching back to 470xx.

ualkyire · 2022-04-22 16:52:16

Ok good news.

"glxinfo" works and "glxgears" work again for the intel iGPU.

The following however don't work:

prime-run glxinfo -B
__GLX_VENDOR_LIBRARY_NAME=nvidia prime-run glxinfo -B
__NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxinfo -B

and they all exit with:

name of display: :0
X Error of failed request:  BadValue (integer parameter out of range for operation)
  Major opcode of failed request:  152 (GLX)
  Minor opcode of failed request:  24 (X_GLXCreateNewContext)
  Value in failed request:  0x0
  Serial number of failed request:  50
  Current serial number in output stream:  51

seth · 2022-04-22 17:00:19

Updated xorg log?

Last edited by seth (2022-04-22 17:00:35)

ualkyire · 2022-04-22 17:04:05

Here's the updated log.

ualkyire · 2022-04-22 17:07:25

I'm noticing this, even though I have acpid installed and enabled as a service:

[    36.962] (WW) Open ACPI failed (/var/run/acpid.socket) (No such file or directory)

I'm also noticing this:

[    38.092] (EE) NVIDIA: Failed to initialize the NVIDIA kernel module. Please see the
[    38.092] (EE) NVIDIA:     system's kernel log for additional error messages and
[    38.092] (EE) NVIDIA:     consult the NVIDIA README for details.

What could be the problem?

Note:
- I created new files in "/etc/modprobe.d" to provide parameters for the modules i915 and nvidia-drm:

/etc/modprobe.d/i915.conf
options i915 enable_dc=1 enable_fbc=1 fastboot=1 modeset=1

/etc/modprobe.d/nvidia-drm.conf
options nvidia-drm modeset=1

I modified "/etc/mkinitcpio.conf" and added:

MODULES=(i915 nvidia nvidia_modeset nvidia_uvm nvidia_drm)

And updated the initramfs.

Any ideas?

EDIT: Just checked I removed "acpid" and forgot to install it again which is why I got the warning above. Do I really need "acpid"?

Last edited by ualkyire (2022-04-22 17:18:06)

ualkyire · 2022-04-22 17:24:06

Sigh just checked "dmesg" and found out that:

[    2.508658] nvidia: loading out-of-tree module taints kernel.
[    2.508667] nvidia: module license 'NVIDIA' taints kernel.
[    2.508668] Disabling lock debugging due to kernel taint
[    2.513956] nvidia: module verification failed: signature and/or required key missing - tainting kernel
[    2.524899] nvidia-nvlink: Nvlink Core is being initialized, major device number 237
[    2.525278] NVRM: loading NVIDIA UNIX x86_64 Kernel Module  390.147  Mon Dec 13 13:56:49 UTC 2021 (using threaded interrupts)
[    2.527320] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  390.147  Mon Dec 13 13:51:26 UTC 2021
[    2.537081] nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 236
[    2.538122] [drm] [nvidia-drm] [GPU ID 0x00000100] Loading driver
[    2.552589] resource sanity check: requesting [mem 0x000e0000-0x000fffff], which spans more than PCI Bus 0000:00 [mem 0x000d0000-0x000e7fff window]
[    2.552592] caller _nv027867rm+0x58/0x90 [nvidia] mapping multiple BARs
[    2.554179] ACPI Warning: \_SB.PCI0.P0P2.VGA._DSM: Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] (20211217/nsarguments-61)
[    2.957708] NVRM: GPU at PCI:0000:01:00: GPU-b6cfc8d5-b7dc-01ba-dac9-9c949ee95365
[    2.957713] NVRM: Xid (PCI:0000:01:00): 31, Ch 00000001, engmask 00000101, intr 50000000
[    2.961462] [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:01:00.0 on minor 1
...
[  172.459992] NVRM: API mismatch: the client has the version 470.103.01, but
               NVRM: this kernel module has the version 390.147.  Please
               NVRM: make sure that this kernel module and all NVIDIA driver
               NVRM: components have the same version.

I'll do proper cleaning of the old dkms 390xx and report back... Sigh...

ualkyire · 2022-04-22 18:03:09

Ok reporting back.

Removed every trace of 390xx, now the 470xx is installed.

Running "prime-run glxinfo" works! The command starts and gets close to finishing, it shows that the OpenGL renderer is the NVIDIA card, but before it completes, KDE starts to lag then freeze, then I get logged out and SDDM no longer works.

Running "journalctl -u sddm" shows "Failed to read display number from pipe" and I have to restart my lap to login again.

I tried "mesa-amber" with the 470xx while using __GLX_VENDOR_LIBRARY_NAME=amber in /etc/profile but it's not working as before.

I installed acpid and enabled its service again, now dmesg looks fine, but Xorg.0.log shows:

[   411.262] (EE) NVIDIA(G0): The NVIDIA X driver has encountered an error; attempting to
[   411.262] (EE) NVIDIA(G0):     recover...
[   411.295] (II) NVIDIA(G0): Error recovery was successful.
[   423.470] (EE) NVIDIA(G0): The NVIDIA X driver has encountered an error; attempting to
[   423.470] (EE) NVIDIA(G0):     recover...
[   423.482] (II) NVIDIA(G0): Error recovery was successful.
[   434.622] (EE) NVIDIA(G0): The NVIDIA X driver has encountered an error; attempting to
[   434.622] (EE) NVIDIA(G0):     recover...
[   434.652] (II) NVIDIA(G0): Error recovery was successful.
[   442.759] (EE) NVIDIA(G0): The NVIDIA X driver has encountered an error; attempting to
[   442.759] (EE) NVIDIA(G0):     recover...
[   442.793] (II) NVIDIA(G0): Error recovery was successful.
[   454.942] (EE) NVIDIA(G0): The NVIDIA X driver has encountered an error; attempting to
[   454.942] (EE) NVIDIA(G0):     recover...
[   454.972] (II) NVIDIA(G0): Error recovery was successful.
[   466.116] (EE) NVIDIA(G0): The NVIDIA X driver has encountered an error; attempting to
[   466.116] (EE) NVIDIA(G0):     recover...

I'm really getting confused here, nothing is working.....

ualkyire · 2022-04-22 18:23:22

I feel like I'm back to square one.

"prime-run" either works with glxinfo or doesn't and "prime-run glxgears" shows a black window with no gears in it and then errors out with:

/usr/bin/prime-run: line 2:  1514 Floating point exception(core dumped) __NV_PRIME_RENDER_OFFLOAD=1 __VK_LAYER_NV_optimus=NVIDIA_only __GLX_VENDOR_LIBRARY_NAME=nvidia "$@"

Found another additional line in Xorg.0.log:

Failed to idle DMA.

Last edited by ualkyire (2022-04-22 18:26:05)

ualkyire · 2022-04-23 08:54:38

Is there something else I could try?

Also, is there any other distribution that might have working versions of the drivers for my cards, until I do a reinstall of arch?

ualkyire · 2022-04-23 10:59:20

Ok I switched to nouveau, and the card is in a decent state, surprised actually that it runs so good (not that I'm expecting much from this card, but I really thought the card was dead lol)!

Screw NVIDIA...

seth · 2022-04-23 13:50:06

Sorry, I didn't receive a mail notification for post #56-#60

[   411.262] (EE) NVIDIA(G0): The NVIDIA X driver has encountered an error; attempting to
[   411.262] (EE) NVIDIA(G0):     recover...

You would again want to compare that to the kernel module messages (dmesg) and possibly try to bbswitch on the GPU first.

Ok I switched to nouveau, and the card is in a decent state … I really thought the card was dead

This might actually be the case, or at least undervoltage.
https://wiki.archlinux.org/title/Nouvea … management
Have you tried running the system on the nvidia GPU alone to ensure its principal functionality (setting aside all optimus hybrid vectors)?

ualkyire · 2022-04-23 18:39:09

seth wrote:

Sorry, I didn't receive a mail notification for post #56-#60

No worries, thanks for all the help.

I tried all possible combinations and everything listed in the wiki, but the outcome was the same.

In the end I stuck with a combination of bbswitch (without bumblebee) and nouveau (with nouveau-fw installed), and things have been really good thus far.

seth · 2022-04-23 19:31:29

The point of the wiki link was to illustrate that maybe the GPU fails to move to higher power modes for physical reasons which is not going to be exposed by nouveau since it'll remain in low power mode.

geopardo · 2022-04-24 00:17:09

The generic kernel.org video driver .. for the lower gpu and the medium gpu are useful .. and efficient
Same problem some time ago .. it said that the xorg server was not reachable.

try... a fresh install........
$ sudo nano /etc/mkinitcpio.conf ADD --> MODULES=(nvidia nvidia-uvm nvidia-modeset nvidia-drm nvidia-peermem)
$ sudo mkinitcpio -P

nvidia kernel parameters settings 
$ sudo nano /etc/default/grub ADD --> GRUB_CMDLINE_LINUX_DEFAULT="loglevel=3 quiet nvidia-drm.modeset=1"
$ sudo grub-mkconfig -o /boot/grub/grub.cfg

xnvidia settings..
$ sudo touch /etc/X11/xorg.conf.d/20-nvidia.conf
$ sudo nano /etc/X11/xorg.conf.d/20-nvidia.conf

touch conf file..
# cat /etc/X11/xorg.conf.d/20-nvidia.conf
Section "Device"
        Identifier "Nvidia Card"
        Driver "nvidia"
        VendorName "NVIDIA Corporation"
        BoardName "GK208B [GeForce GT 710] (rev a1)"
        Option "ConnectedMonitor" "DFP"
        Option "UseEDID" "True"
        Option "UseEDIDFreqs" "True"
        Option "UseEDIDDpi" "True"
        Option "DPI" "75 x 75"
        Option "AllowIndirectGLXProtocol" "True"
EndSection[/forum]

nvidia settings..
$ sudo pacman -S nvidia-settings
$ nvidia-settings -q CurrentMetaMode

hardware accelerated video encoding...
$ sudo touch /etc/udev/rules.d/70-nvidia.rules
$ sudo nano /etc/udev/rules.d/70-nvidia.rules

$ cat /etc/udev/rules.d/70-nvidia.rules
ACTION=="add", DEVPATH=="/bus/pci/drivers/nvidia", RUN+="/usr/bin/nvidia-modprobe -c0 -u"
status...
$ dkms status
$ reboot
$ lsmod
$ sudo dmesg | grep nvidia
$ sudo mkinitcpio -MTest

Last edited by geopardo (2022-04-25 19:26:24)

seth · 2022-04-24 06:32:56

1. there's no "generic kernel.org video driver", but I guess you mean nouveau
2. the subsequently constantly referenced nvidia binary blob is certainly not part of that
3. you don't seem to have an optimus system, which is the main problem of this thread
4. you have a Kepler card and can't use "nvidia-settings", bu need nvidia-470xx-settings
5. use code tags, https://bbs.archlinux.org/help.php#bbcode

@ualkyire before you attempt to do any of #66, click the users name and check out his (only) other topic.
Then make up your mind on whether you want to pay any attention to his comments…

ualkyire · 2022-04-28 17:40:14

@seth thanks for pointing that out.

Should I mark this as solved, even though it kinda isn't?

seth · 2022-04-28 20:42:03

It's your thread.
If you'd prefer to understand the issue and ultiamtely use the nvidia blob, I'd first look at

seth wrote:

Have you tried running the system on the nvidia GPU alone to ensure its principal functionality (setting aside all optimus hybrid vectors)?

ualkyire · 2022-04-28 20:58:07

seth wrote:

Have you tried running the system on the nvidia GPU alone to ensure its principal functionality (setting aside all optimus hybrid vectors)?

Yup, I have, to no avail.

I tried using the configuration mentioned in the wiki about running the NVIDIA card alone (while removing all the optimus stuff), I then tried all of the available optimus options with options to use the card alone also to no avail. The problem always occurs when X is starting which lead me to believe that something is weirdly wrong with how Xorg autodetects/configures itself, even with tens of different configurations tried, nothing worked...

Weirdly enough, nouveau for kepler cards, is pretty decent, I think reclocking also works for this generation of cards, and the temps are good. The only bad thing is the power consumption, but with bbswitch, it's back to normal as if the card is off.

How I look at things now, I got my optimus setup working with little hassle with nouveau/nouveau-fw/bbswitch, the temperatures are good, energy consumption is great thanks to bbswitch, and performance is atleast 60-70% of the original NVIDIA driver (at least for what I'm using), the card also switches instantly (less latency compared to other setups), so I'm happy with this configuration.

vegna · 2022-04-29 07:34:49

Hi ualkyirie,

I "solved" the issue that prevented the nouveau driver loading in integrated mode:
It seems like that if you download software like nvclock and nvtop they call the nvidia driver continously "blacklisting" the nouveau one.

 DRI_PRIME=1 glxinfo | grep "OpenGL renderer"
OpenGL renderer string: NV108

DRI_PRIME=2 glxinfo | grep "OpenGL renderer"
OpenGL renderer string: Mesa Intel(R) HD Graphics 4600 (HSW GT2)

I can't say anything on the power saving, I'm working from home and i don't have that issue.

Arch Linux

#51 2022-04-22 15:02:44

Re: NVIDIA Optimus isn't working for 470xx

#52 2022-04-22 15:14:24

Re: NVIDIA Optimus isn't working for 470xx

#53 2022-04-22 15:15:56

Re: NVIDIA Optimus isn't working for 470xx

#54 2022-04-22 16:52:16

Re: NVIDIA Optimus isn't working for 470xx

#55 2022-04-22 17:00:19

Re: NVIDIA Optimus isn't working for 470xx

#56 2022-04-22 17:04:05

Re: NVIDIA Optimus isn't working for 470xx

#57 2022-04-22 17:07:25

Re: NVIDIA Optimus isn't working for 470xx

#58 2022-04-22 17:24:06

Re: NVIDIA Optimus isn't working for 470xx

#59 2022-04-22 18:03:09

Re: NVIDIA Optimus isn't working for 470xx

#60 2022-04-22 18:23:22

Re: NVIDIA Optimus isn't working for 470xx

#61 2022-04-23 08:54:38

Re: NVIDIA Optimus isn't working for 470xx

#62 2022-04-23 10:59:20

Re: NVIDIA Optimus isn't working for 470xx

#63 2022-04-23 13:50:06

Re: NVIDIA Optimus isn't working for 470xx

#64 2022-04-23 18:39:09

Re: NVIDIA Optimus isn't working for 470xx

#65 2022-04-23 19:31:29

Re: NVIDIA Optimus isn't working for 470xx

#66 2022-04-24 00:17:09

Re: NVIDIA Optimus isn't working for 470xx

#67 2022-04-24 06:32:56

Re: NVIDIA Optimus isn't working for 470xx

#68 2022-04-28 17:40:14

Re: NVIDIA Optimus isn't working for 470xx

#69 2022-04-28 20:42:03

Re: NVIDIA Optimus isn't working for 470xx

#70 2022-04-28 20:58:07

Re: NVIDIA Optimus isn't working for 470xx

#71 2022-04-29 07:34:49

Re: NVIDIA Optimus isn't working for 470xx

Board footer