OPINIONATED OMARCHY
TROUBLESHOOTING CORPUS
790
RECORDS
790
AUDITED
0
UNCHECKED
2388
SOURCES
LIVE
audio-input high occasional DISPUTE

Choose between SOF and legacy HDA when speakers or the internal mic stay dead

omarchyarchintellaptopsofpipewirelimine
AUDITED Audited against its sources and confirmed accurate. · checked against Omarchy omarchy 4.0.4-1 on 2026-10-05
AUDIT NOTE

Kernel source (sound/hda/core/intel-dsp-config.c via gh api) confirms MODULE_PARM_DESC '0=auto, 1=legacy, 2=SST, 3=SOF, 4=AVS', the exact 'SoundWire enabled on CannonLake+ platform, using SOF driver' and 'Digital mics found on Skylake+ platform, using SOF driver' strings, the PCI-class then NHLT DMIC/SoundWire decision, and that a nonzero dsp_driver is returned directly (except on HSW/BDW/BYT/BSW, where it is ignored). The parameter is 0444, so /sys/module/snd_intel_dspcfg/parameters/dsp_driver is readable. snd-intel-dspcfg.ko.zst is a loadable module in both /lib/modules/7.2.5-3-omarchy and 7.2.3-arch1-3 on this machine. thesofproject/sof#11117 matches the record exactly (K3605ZF ALC256: dsp_driver=1 loses the DMIC, dsp_driver=3 loses speakers), and a maintainer there suspects UCM rather than the codec driver, which the record's cause already allows for. omacom/omarchy#2332 comments confirm dsp_driver=1 fixed dead speakers under SOF on Huawei laptops. Omarchy 4 branch checked locally: limine-mkinitcpio and limine-update exist (limine-mkinitcpio-hook 1.38.0), limine-update runs limine-install then limine-mkinitcpio, the KERNEL_CMDLINE[default]+= drop-in form matches /etc/limine-entry-tool.d/omarchy-defaults.conf, and /etc/default/limine also uses += so a drop-in is not overridden. The initramfs rebuild is unnecessary for a sound module but harmless and correct. The Arch ALSA wiki page no longer mentions dsp_driver and supports only the sof-firmware context. Nothing was exercised on audio hardware.

RISK

dsp_driver=1 removes the internal digital microphone on DMIC laptops, and a mode that does not suit the hardware can leave no working sound card at all. Booting is not affected, so the revert is always available: delete /etc/modprobe.d/dsp-driver.conf (or the /etc/limine-entry-tool.d/99-dsp-driver.conf drop-in), rebuild as above and reboot.

Symptom

Intel laptop with sof-firmware already installed, and still broken in one of two ways. Either the speakers are silent although the SOF card is present (Dummy Output, or a sink that plays nothing), or the speakers work but the built-in microphone is missing entirely. journalctl -k -b shows Digital mics found on Skylake+ platform, using SOF driver or SoundWire enabled on CannonLake+ platform, using SOF driver. A forum fix of dsp_driver=1 makes the speakers work but the internal mic vanishes.

Cause

The kernel module snd_intel_dspcfg decides which driver owns the Intel audio DSP. In automatic mode it reads the PCI class and the ACPI NHLT table, and when it finds digital microphones (or SoundWire) it hands the device to SOF. SOF then needs a topology and a UCM profile matching the codec. Where none fits, for example the ASUS Vivobook K3605ZF with an ALC256, SOF captures the DMICs fine but the speakers produce nothing. The module parameter dsp_driver overrides the choice (0=auto, 1=legacy, 2=SST, 3=SOF, 4=AVS, from MODULE_PARM_DESC in sound/hda/core/intel-dsp-config.c). Legacy HDA (snd_hda_intel) drives the analog codec but cannot reach microphones wired to the DSP, which is why the internal mic disappears with dsp_driver=1. This is a different failure from missing SOF firmware, where the card never appears at all. snd-intel-dspcfg is a loadable module in both linux-omarchy 7.2.5-3 and stock linux (checked in /lib/modules/7.2.5-3-omarchy/kernel/sound/hda/core/).

Fix

**1. See which driver was chosen and why.**


journalctl -k -b | grep -iE 'using SOF driver|using SST driver|dsp_driver|sof-audio'
lspci -k | grep -A3 -i audio            # Kernel driver in use: sof-audio-pci-intel-* or snd_hda_intel
cat /sys/module/snd_intel_dspcfg/parameters/dsp_driver

**2. Try legacy HDA if the speakers are dead under SOF.**


echo 'options snd_intel_dspcfg dsp_driver=1' | sudo tee /etc/modprobe.d/dsp-driver.conf

Rebuild the initramfs so the copy that the modconf hook bundles agrees with /etc:


# Omarchy 4 (Limine UKI)
sudo limine-mkinitcpio
# plain Arch
sudo mkinitcpio -P

Reboot. Expect working speakers and headphone jack and no internal digital mic. A USB or Bluetooth headset mic still works.

**3. Or force SOF if automatic mode picked legacy and the internal mic is missing.** This needs sof-firmware:


echo 'options snd_intel_dspcfg dsp_driver=3' | sudo tee /etc/modprobe.d/dsp-driver.conf

Then rebuild and reboot as above.

**Kernel command line instead of modprobe.d**, if you prefer:


# Omarchy 4: the command line lives inside the UKI, set it with a drop-in
printf 'KERNEL_CMDLINE[default]+=" snd_intel_dspcfg.dsp_driver=1"\n' \
  | sudo tee /etc/limine-entry-tool.d/99-dsp-driver.conf
sudo limine-update

Plain Arch: append snd_intel_dspcfg.dsp_driver=1 to GRUB_CMDLINE_LINUX_DEFAULT in /etc/default/grub and run sudo grub-mkconfig -o /boot/grub/grub.cfg, or to the options line of your /boot/loader/entries/*.conf with systemd-boot.

If neither mode gives both speakers and mic, the machine needs a kernel quirk or a UCM profile that does not exist yet. Report it to the SOF project with alsa-info output rather than stacking more module options.

Verify

After reboot cat /sys/module/snd_intel_dspcfg/parameters/dsp_driver shows the value you set, lspci -k shows the matching driver in use, and speaker-test -c 2 -t wav plays. Check wpctl status for which sources exist so you know whether the internal mic survived.

Sources

intel-sof-vs-legacy-hda-dsp-driver-tradeoff