IO board is the same as Quanta AM9A IO, idiot put ferrite bead inline on speaker output - i removed them and bridged with wire so this gives nice clean sound without white noise/ hiss

the remaining problem is to disable power save (unsure if it can be reliably done with maxxaudio software, none of the registry tweaks worked and with the one involving PerformanceIdleTime the value kept resetting on restart)

compared audio signals coming from CPU, such as ACZ_SDOUT_R ACZ_BCLK_R etc. on this platform with the older HP compaq platform. in the compaq these signals are bypassed with small value cap to GND (there are 0ohm resistors inline as well) while in quanta all the signals (except HDA_SDIN0) are pulled to GND through 1K resistor.

and while audio output to speakers shuts down when no any kind of sounds, the mic input that comes to HDA_SDIN0 stays on all the time

are these resistors a way to disable power gating ?

before messing with mainboard i took a closer look at io board, more specifically NB_MUTE# EAPD# SUB_MUTE# pins. taped the two mute pins on connector to break them from EC, ALC still generates 3.3V on NB_MUTE - sound played fine
also tried taking out BAT54, this generated some extra pops on sub and helped to confirm HKEY_CURRENT_USER\Software\Realtek\Audio\RtkNGUI64\PowerMgnt settings actually do work but apply only to subwoofer output
after putting diode back i tried to set DelayTime to 0 while Enabled still at 1. everytime the sound is played and stopped there is a pop on subwoofer, along with faint pop on speakers (maybe adding discharge resistors on the other LDO pins could help, or add bigger value on LDO that already has it)
when DelayTime is bigger sub has enough time to power off before pop hits it, so it looks like they used power gating to mask sloppy and unfinished design
remains to see if possible to force PerformanceIdleTime to anything other than 0, i feel like it should configure speakers LDO