Greetings to all.
I have a rather peculiar issue that I'm wondering if someone can shed some light on.
I'm running a dual E5-2690v4 setup with an Asus Z10PA-D8 motherboard and a Creative AE-7 sound card. I'm currently running Pipewire (with the relevant pulseaudio daemons masked) with Wireplumber on a systemd based stage3 install.
Everything seems to work great, until the optical SPDIF output on my AE-7 just... stops. This almost always happens mid-stream while something is playing, so it's not an issue with output suspension. pw-top still reports that audio output should be happening, and `cat /proc/asound/Creative/pcm1p/sub0` seems to confirm that the kernel thinks it's still outputting to the SPDIF port. There are no entries in journalctl or dmesg that would suggest anything weird is going on (nothing like an underflow/overflow warning).
Other things I've tried:
- Restricting the output of Pipewire to 48kHz (which AFAIK is the default)
- Monkeying around with the quantum default/min/max sizes
- Disabling session suspension just for the hell of it
- Changing api.alsa.period-size to 1024
- Changing api.alsa.headroom to 8192
- Recompiling for CONFIG_PREEMPT (formerly on voluntary)
- Trying different versions of the card's ctefx.bin firmware file
- Disabling NUMA balancing via /proc/sys/kernel/numa_balancing to see if it's some weird dual CPU voodoo thing
- Updating wireplumber and pipewire to ~amd64
I can get the port working again by simply changing the audio output to analogue then back again to SPDIF and telling whatever was playing (usually Fooyin or Firefox) to start playing again. Things will continue to work for anywhere from 15 minutes to 6 hours, until SPDIF just... heads off into the sunset again, and my speakers go dead.
I'm at a complete loss as to what to troubleshoot next. This never happened under Windows, but then again I never upgraded past Windows 10 and I was using an ancient copy of the Creative drivers from 2020 that I never bothered updating. Everything works fine until it doesn't and I literally don't have any complaints otherwise.
The only thing I can personally think of doing is ditching the AE-7 and buying something a bit less proprietary, but I was kind of hoping not to do that until I'd exhausted all other avenues of troubleshooting with the AE-7 itself.
Cheers,
-CMPX
Creative AE-7 randomly cutting out with SPDIF mid-stream
Message
In case anyone else is having the same problem...
My particular issue appears to have been caused by the fact the sound card is installed in PCI-e slot 1, which is located behind the system PCH chipset and can introduce minor delays when communicating with the card if the PCI-e bus is sufficiently loaded down. Apparently the CA0132 codecs are VERY sensitive to DMA timing and if *anything* goes wrong, the card will simply stall and halt output. I would have noticed this if I'd bothered to debug anything with the 3.5mm ports instead of SPDIF, since it's quite obvious that something is going wrong when using 3.5mm since the output status in /proc/asound/Creative/*/sub0/status will show an avail_max of 0 with the hw_ptr value refusing to increase (indicating a hung codec).
The solution for my system was to add the following to /etc/modprobe.d/creative.conf (or whatever file you want really):
options snd_hda_intel bdl_pos_adj=-1,-1,64
In my case the first two values (-1) are for an ALC892 codec and an AMD 7900 XTX HDMI output respectively, with the Soundblaster AE-7 as the third card (according to the indicates listed by /proc/asound/cards). If you only have a single AE-7 and no other codecs installed in the system, then that'd just be bdl_pos_adj=64 and nothing else.
This appears to have fixed the problem entirely. Output is now rock solid regardless of how the system is being used.
-CMPX
My particular issue appears to have been caused by the fact the sound card is installed in PCI-e slot 1, which is located behind the system PCH chipset and can introduce minor delays when communicating with the card if the PCI-e bus is sufficiently loaded down. Apparently the CA0132 codecs are VERY sensitive to DMA timing and if *anything* goes wrong, the card will simply stall and halt output. I would have noticed this if I'd bothered to debug anything with the 3.5mm ports instead of SPDIF, since it's quite obvious that something is going wrong when using 3.5mm since the output status in /proc/asound/Creative/*/sub0/status will show an avail_max of 0 with the hw_ptr value refusing to increase (indicating a hung codec).
The solution for my system was to add the following to /etc/modprobe.d/creative.conf (or whatever file you want really):
options snd_hda_intel bdl_pos_adj=-1,-1,64
In my case the first two values (-1) are for an ALC892 codec and an AMD 7900 XTX HDMI output respectively, with the Soundblaster AE-7 as the third card (according to the indicates listed by /proc/asound/cards). If you only have a single AE-7 and no other codecs installed in the system, then that'd just be bdl_pos_adj=64 and nothing else.
This appears to have fixed the problem entirely. Output is now rock solid regardless of how the system is being used.
-CMPX
- flexibeast
- l33t
- Posts: 682
- Joined: Mon Apr 04, 2022 4:15 am
- Location: Naarm/Melbourne, Australia
- Contact:
Sadly it doesn't seem to have solved the problem entirely.
About an hour after I posted that it cut out again... It seems like bdl_pos_adj greatly reduced the frequency of the issue, but didn't eliminate it. I tried jacking it to 128 afterwards but it cut out a couple hours after that. Also moved the card to a different slot (one that's not behind the PCH) but that didn't make a difference either.
At this point I'm at a complete and utter loss as to what is going on. It's always the same behavior- SPDIF drops with no indication anything is wrong, or if you're using 3.5mm then you can clearly see the codec die in /proc/asound. I'm starting to think these are just exceptionally poorly engineered cards in general since I remember having problems under Windows until I installed the official Creative drivers (which heavily suggests they're doing some sort of unknown workarounds in software to keep the codec running smoothly).
Guess I'm in the market for a replacement PCI-e sound card.
-CMPX
About an hour after I posted that it cut out again... It seems like bdl_pos_adj greatly reduced the frequency of the issue, but didn't eliminate it. I tried jacking it to 128 afterwards but it cut out a couple hours after that. Also moved the card to a different slot (one that's not behind the PCH) but that didn't make a difference either.
At this point I'm at a complete and utter loss as to what is going on. It's always the same behavior- SPDIF drops with no indication anything is wrong, or if you're using 3.5mm then you can clearly see the codec die in /proc/asound. I'm starting to think these are just exceptionally poorly engineered cards in general since I remember having problems under Windows until I installed the official Creative drivers (which heavily suggests they're doing some sort of unknown workarounds in software to keep the codec running smoothly).
Guess I'm in the market for a replacement PCI-e sound card.
-CMPX