Re: [fluid-dev] fluidsynth with realtime kernel

[Pedro Lopez-Cabanillas]

On Monday 11 October 2010, you wrote:
> On 2010-10-08 23:15, Pedro Lopez-Cabanillas wrote:
> > Anyway, even if you use the same operating system, kernel version, 
priority
> > settings, environment, processor, etc. a critical factor is the audio 
hardware
> > device. Professional or good quality audio interfaces typically allow much
> > smaller buffers, and lower latency as a consequence than cheap/low quality
> > devices. If your laptop has an HDA sound hardware, it is probably the 
culprit.
> 
> Do you know if there are any specific HDA controllers and/or codecs that 
> work better or worse in this regard? I'd be very interesting to know.

I also would like to know the complete list of HDA hardware implementations to 
avoid, but I can only say that my Asus laptop has one, and it is of the bad 
quality category. I've seen other laptops suffering from the same problems. 
FS+alsa needs 3x512 buffers, which is the minimum acceptable to avoid xruns.

> My experience differs from yours - on my HDA here, I can run with 
> driver=alsa, 4 x 64 buffers quite stable (as in, if there are xruns, I 
> can't hear them). Something I consider OK, especially given the fact 
> that I don't run any special rt kernel (just Ubuntu's standard one, 
> called "generic"), and haven't done the limits.conf stuff either.

My Mac Mini also has HDA, and the sound is very good with little latency. I've 
tested a few computers with mobo-integrated HDA, also with good results. The 
only rule seems to be that many things are called "Intel HDA", with a very 
wide range of quality, from quite good to very poor. Something like that also 
happened with AC'97 variants, but in general wasn't so terrible.

Regards,
Pedro