Re: pulseaudio / alc1220 sound clicks interruptions etc

From: Tomek CEDRO <tomek_at_cedro.info>
Date: Thu, 05 Jun 2025 17:06:00 UTC
On Thu, Jun 5, 2025 at 5:58 PM Amar Takhar <verm@darkbeer.org> wrote:
> On 2025-06-05 17:34 +0200, Tomek CEDRO wrote:
> > Yeah, I think problem is somewhere below, audio and gpu only gets hit
> > by the problem. Even with PCI-Ex16 set to 8x/8x and all C-states
> > disabled I just got all onboard USB controllers fail. I have
> > additional PCI-Ex1 USB3.0 controller that saves me from that and the
> > machine works no kernel panic. Maybe we should try 8x/4x/4x PCI mode?
>
> I've been dealing with this for 3+ years now and I've tried all of these every
> once in a while I'd get annoyed and investigate more hoping things have changed.
> Changing the PCI mode won't help it will just cripple your machine and make the
> problem less obvious but not remove it.  I had to run for 8+ months without E
> cores until #261169 was fixed.
>
> It's most likely something to do with the scheduler though it could be something
> in the audio stack that won't help the other strange hiccups and slowdowns and
> the audio issues may get fixed if those are as well.

This AlderLake from Intel is very interesting because although it
provides really strong computing resources (not to mention NPU and
built-in GPU) it also has very good dynamic power management built
into the hardware - being ~10x faster than my old CPU it takes 1/2 of
the standard use power and only 1/5 more power when fully loaded. Thus
my decision to get Intel Ultra9 over AMD Zen5.

AlderLake has this built-in features like dynamic frequency scaling
per core, P and E cores, resource priorities, many more types and
classes of C-states than before, ACPI and WMI, etc etc. Hardware can
put to sleep cores that are not 100% utilized, maybe even passing
tasks from core to core, which may not be handled properly and causing
the problems we experience (delays, hangs, glitches, peripheral
issues, panics, etc). I think this needs separate and dedicated
handling by the kernel even below scheduler.

Could Intel get interest in FreeBSD as the consumer base? I tried
several times but failed miserably :-(

-- 
CeDeROM, SQ7MHZ, http://www.tomek.cedro.info