[linux-audio-user] Real-time kernel

Paul Davis paul at linuxaudiosystems.com
Thu Dec 21 20:47:37 EST 2006

On Thu, 2006-12-21 at 16:47 -0800, Ken Restivo wrote:
> Hash: SHA1
> That's bleak.
> I bought a Mac Mini to do audio on. It's got the dreaded hda_intel chip. 
> But with kernel 2.4.18 it works fine. Fine meaning: there's still some xruns, but not enough to slow me down too much.
> Right now I haven't been able to get it to actually *record*, but since I'm mostly doing all-digital stuff right now it hasn't gotten in my way. Yet.
> I suppose if I do any serious work I'll get one of those firewire audio external boxes like the FA-66, which aren't too expensive last I checked.

no experience with any one HDA setup can be transferred to any other.

the intel HDA specification is barely a real specification. it has left
enough "wiggle room" that the drivers for "HDA" are actually a
hodge-podge (technical term) of random fixes to patch up areas left
insufficiently defined.

there are dozens of HDA configurations - each one of them is unique. and
tomorrow, when apple, or HP, or sony or fujitsu or whoever release a new
laptop, there will likely be one more.

this works on linux because the h/w manufacturer hacks the base driver
supplied by intel and includes it in the windows kit shipped with the
laptop/desktop. on linux, they do nothing (of course) so it comes down
to the user finding out that it doesn't work, and having the
guts/know-how to track down what register mappings have been changed,
which pinouts have been reordered and then trying to get a new quirk/fix
added to the HDA driver.

its deeply pathetic, and i think that intel should be absolutely ashamed
of what they have done. even creative did a better job than this,
without even trying to: the SB16 was a standard; not very good, but a
standard nevertheless. 


More information about the Linux-audio-user mailing list