At Mon, 22 Nov 2004 10:30:21 -0500,
Eric Sharkey wrote:
> 
> 
> Hi all,
> 
> I posted on alsa-user a while back, but I think I need developer
> assistance with this.
> 
> There seems to be a problem with Alsa when running on the AMD64
> architecture on motherboards with the Via K8T800 chipset.  The sound
> is highly irregular, with lots of drop-outs, but also speed-ups,
> slow-downs and weird volume changes.
> 
> I've got this problem on an Asus K8V SE motherboard.  Rod Smith
> has the same problem on an MSI Neo-FSR.
> (http://groups.google.com/groups?q=%22asus+k8v%22+alsa&hl=en&lr=&c2coff=1&selm=1et59c-90v.ln%40speaker.rodsbooks.com&rnum=4)
> 
> In that post, Rod thought the problem was the Alsa driver for the
> on-board sound (VIA VT8237), but this is not the case, as I've installed
> a PCI Trident 4DWave NX, and it shows exactly the same behavior.
> The problem appears not to be in the low level driver code, but
> something higher.
> 
> The degree of the problem is highly sensitive to the load on the
> CPU at the time.  Games like bumprace, which use multiple threads
> and never sleep (giving load values around 8), sound awful.  Most
> games like tuxkart, which keep the load under 1, sound perfectly fine.
> 
> And yet, some things sound bad even when the CPU isn't loaded.
> timidity++ is a good example.

The current trident driver seems buggy with 2 periods.
The patch below is a workwround (not a right fix, though).


> This happens with both 64 and 32 bit kernels, and no amount of
> twiddling with kernel parameters (ACPI/CPU frequency scaling, apic,
> preemption, etc.) seems to make any difference.

Hmm, I just thought of cpurfreq, but your report sounds that it's not
the case.  Could you unload powernow-k8 module just to be sure?

I think you'd better to ask this on LKML, since the symptom looks like
the kernel core problem.


Takashi