At Wed, 23 Oct 2002 15:16:08 +0200, Werner Schweer wrote: > > On Wednesday 23 October 2002 11:51, Takashi Iwai wrote: > ... > > > - SB AWE models (ugh, crap!) > > > - Yamaha YMF7xx/DS-XG (some have reported that these work ok, > > > but in any case they have a max 3 periods limitation > > > similar to cs4281, which can confuse apps) > > > > no, instead, the interrupts are generated in the fixed time-length, > > not at the period boundary or the end of buffer. > > thus, this chip doesn't suit for low-latency purpose at all. > > > > the similar case is ESS chips, es1968 and maestro3 (allegro). > > that is, many on-board chips on notebooks are crap, unfortunately. > > it cost me a day of debugging to find out that my notebook with > ES1983S (Maestro3i) chip does not work well with JACK. Sometimes > all seems to be ok (no underruns) sometimes timing was horrible with > lots of underruns. Tracing shows that soundcard interrupts (JACK callbacks) > sometimes are just too late. > Takashi, are you saying that this cannot be fixed because the driver > does not get an interrupt on buffer empty?
i'm afraid that it's a driver bug. i got the similar report from Steve, and in fact, this delay looks longer than expected. however, in the case of maestro3, the situation is worse: we have no enough technical information to debug this behavior... regarding to the audio device for notebooks, a USB device is not a bad choice nowadays. the latency is fairly good, you'll get 1 or 2ms order response. but the bandwidth is too narrow, so the i/o channels are limited. and if you use other usb devices together, the behavior is unreliable. but, anyway, it's easy - just hook and run (as long as usb hotplug is fixed :) ciao, Takashi
