Audio problems with the 415.18 drivers

+1

Same problem with 415.25, as of Feb 2019.

Display port output, pulse shows all outputs as “unplugged” after upgrade.

2019-02-22 18:37:37 install nvidia-415:amd64 <none> 415.27-0ubuntu0~gpu16.04.1
2019-02-22 18:37:50 install nvidia-opencl-icd-415:amd64 <none> 415.27-0ubuntu0~gpu16.04.1
2019-02-22 18:37:51 configure nvidia-415:amd64 415.27-0ubuntu0~gpu16.04.1 <none>
2019-02-22 18:39:45 status installed nvidia-415:amd64 415.27-0ubuntu0~gpu16.04.1
2019-02-22 18:39:45 configure nvidia-opencl-icd-415:amd64 415.27-0ubuntu0~gpu16.04.1 <none>
2019-02-22 18:39:45 status installed nvidia-opencl-icd-415:amd64 415.27-0ubuntu0~gpu16.04.1

Also broken under “NVIDIA-Linux-x86_64-418.43.run” freshly downloaded and tested. Same issue

/proc/asound/NVidia$ cat eld#0.0
monitor_present		0
eld_valid		0
/proc/asound/NVidia$ cat eld#0.1
monitor_present		0
eld_valid		0

and

[    4.317673] snd_hda_codec_hdmi hdaudioC0D0: HDMI: invalid ELD data byte 0

The type of GPU would be interesting, to see if there’s some kind of pattern. The first ones were
GTX 980
Quadro K2100M
Connectors, adapters, converters used?

Mine is a GTX 970 4096 MB directly connected to a TV via an HDMI cable which works perfectly well with the 390 driver.

GT 1030 - Display port to HDMI cable, connected to Yamaha Amp

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 410.78       Driver Version: 410.78       CUDA Version: 10.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GT 1030     Off  | 00000000:01:00.0  On |                  N/A |
| 35%   34C    P8    N/A /  30W |    149MiB /  2000MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

Still the same after upgrade to

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 415.27       Driver Version: 415.27       CUDA Version: 10.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Quadro K2000M       Off  | 00000000:01:00.0  On |                  N/A |
| N/A   49C    P0    N/A /  N/A |    334MiB /  1999MiB |      2%      Default |
+-------------------------------+----------------------+----------------------+
[  133.040133] snd_hda_codec_hdmi hdaudioC2D0: HDMI: invalid ELD data byte 0

I’m using a Thinkpad laptop with a Hama miniDP to HDMI/DVI/DP adapter. This setup has worked fine for years (so this rules out the adapter as an issue).

nvidia-bug-report.log.gz (1.04 MB)

This is what I see on dmesg if I run:

echo "OFF" > /proc/acpi/nvhda
echo "ON" > /proc/acpi/nvhda
[ 2274.540678] nvhda: Found nv audio device 0000:01:00.1
[ 2274.641277] pci 0000:01:00.1: Dropping the link to 0000:01:00.0
[ 2274.641283] iommu: Removing device 0000:01:00.1 from group 1
[ 2274.641340] nvhda: disabling audio
[ 2279.003518] nvhda: enabling audio
[ 2279.003535] pci 0000:01:00.1: [10de:0e1b] type 00 class 0x040300
[ 2279.003547] pci 0000:01:00.1: reg 0x10: [mem 0xf1080000-0xf1083fff]
[ 2279.003657] iommu: Adding device 0000:01:00.1 to group 1
[ 2279.003700] nvhda: Audio found, adding
[ 2279.003705] pci 0000:01:00.1: BAR 0: assigned [mem 0xf1080000-0xf1083fff]
[ 2279.003719] pci 0000:01:00.1: Linked as a consumer to 0000:01:00.0
[ 2279.003841] snd_hda_intel 0000:01:00.1: Disabling MSI
[ 2279.003845] snd_hda_intel 0000:01:00.1: Handle vga_switcheroo audio client
[ 2279.577601] input: HDA NVidia HDMI/DP,pcm=3 as /devices/pci0000:00/0000:00:01.0/0000:01:00.1/sound/card2/input22
[ 2279.577857] input: HDA NVidia HDMI/DP,pcm=7 as /devices/pci0000:00/0000:00:01.0/0000:01:00.1/sound/card2/input23
[ 2279.578049] input: HDA NVidia HDMI/DP,pcm=8 as /devices/pci0000:00/0000:00:01.0/0000:01:00.1/sound/card2/input24

Just updated to 418.43-4 (arch), still no luck.

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.43       Driver Version: 418.43       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 970     Off  | 00000000:01:00.0  On |                  N/A |
| 28%   37C    P0    48W / 151W |    132MiB /  4039MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

Here

06:00.0 VGA compatible controller: NVIDIA Corporation GP108 [GeForce GT 1030] (rev a1)
07:00.0 VGA compatible controller: NVIDIA Corporation GK106 [GeForce GTX 660] (rev a1)

both with same Problem.

My card is a GTX660. I have also been struggling against the problem since Archlinux upgraded to 415.xx and then since Voidlinux upgraded to 418.xx. But with 418.xx, on Void at least (I am not in the mood to test with Arch/systemd.), the bug seems to affect only the DisplayPort audio output. I have a three monitor setup. One is connected through DisplayPort and two were attached through DVI. I’ve changed one of the DVI connections to HDMI. Surprisingly, the video card recognized the audio output on the HDMI, but still not on the DisplayPort. That is the makeshift I will use for now.

Mine is a quadro k5200.

Mine is a GTX 970. Same issue with having the sound from DP (card) to HDMI (monitor). However, it is not a 100% failure. Maybe 10-20% of the times I boot, the ELD is detected. Monitor is a Samsung UHD TV (Q7FN). So it looks like a race condition issue - at least in my case. Maybe a problem with systemd and the order the drivers are loaded at boot? Any suggestion about a way to pinpoint the problem? I am happy to compare log outputs between failures and success.
OS is Archlinux, the problem happend from drivers 415, drivers 410 were working properly.

Does

sudo hdajacksensetest -s

change anything in hdmi audio detection?
(alsa-tools have to be installed)

No, and I do not think this is relevant. This command only detects the jack plugs, with analog signal.

Pin 0x14 (Green Line Out, Rear side): present = No
Pin 0x15 (Black Line Out, Rear side): present = No
Pin 0x16 (Orange Line Out, Rear side): present = No
Pin 0x18 (Pink Mic, Rear side): present = No
Pin 0x19 (Pink Mic, Front side): present = No
Pin 0x1a (Blue Line In, Rear side): present = No
Pin 0x1b (Green Headphone, Front side): present = No

That’s the onboard sound, use

cat /proc/asound/cards

to find the correct index and then use

sudo hdajacksensetest -s -c X

with X being the found index of the nvidia hda.
Edit: my cat ran away but caught it now.

  1. Thanks for the tip, very useful indeed:
$ sudo hdajacksensetest -s -c 1
Pin 0x04 ( Digital Out, HDMI): present = No
Pin 0x05 ( Digital Out, HDMI): present = Yes
Pin 0x06 ( Digital Out, HDMI): present = No
Pin 0x07 ( Digital Out, HDMI): present = No

The strange thing is that my cable is connected physically to HDMI 1 on the Samsung, not HDMI 2. But it says (when detected) HDMI 2 in pavucontrol.

  1. Now I am wondering: since it looks like a race condition (I have detection every 5 to 10 reboots), I could trace the boot sequence and see if they do differ. With this working boot, I have:
$ dmesg|grep HDA\ NV
[    5.049550] input: HDA NVidia HDMI/DP,pcm=3 as /devices/pci0000:00/0000:00:01.0/0000:01:00.1/sound/card1/input29
[    5.049591] input: HDA NVidia HDMI/DP,pcm=7 as /devices/pci0000:00/0000:00:01.0/0000:01:00.1/sound/card1/input30
[    5.049618] input: HDA NVidia HDMI/DP,pcm=8 as /devices/pci0000:00/0000:00:01.0/0000:01:00.1/sound/card1/input31
[    5.049642] input: HDA NVidia HDMI/DP,pcm=9 as /devices/pci0000:00/0000:00:01.0/0000:01:00.1/sound/card1/input32

but I know that systemd provides better tools to record the boot sequence.

  1. By the way, am I the only one to observe this hit and miss behaviour?

Was that hdajacksensetest output from the non-working state? You could then try to read the ELD using
sudo hda-verb /dev/snd/hwC1D0 0x5 0xf2e 0x8
or
sudo hda-verb /dev/snd/hwC1D0 0x5 0xf2f 0
and see if it changes anything.

$ sudo hdajacksensetest -s -c 1
    Pin 0x04 ( Digital Out, HDMI): present = No
    Pin 0x05 ( Digital Out, HDMI): present = No
    Pin 0x06 ( Digital Out, HDMI): present = No
    Pin 0x07 ( Digital Out, HDMI): present = No

I’ll post the other information soon.

These commands do not change anything:

$ cat /proc/asound/NVidia/eld#0.1
monitor_present         0
eld_valid               0

However I may be onto something. I just restarted and saved the output of dmesg. I can check how the order of entries differ in non working and working cases. Could it be the race condition this problem suggests? I need to do further dmesg dumps to confirm.