On Thu, 2008-01-24 at 08:51 +0100, Florent Berthaut wrote:
  

Fernando Lopez-Lezcano a écrit : 
    

On Wed, 2008-01-23 at 16:44 +0100, Florent Berthaut wrote:
      

Florent Berthaut a écrit : 
        

Florent Berthaut a écrit : 
          

Hartmut Noack a écrit : 
            

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Florent Berthaut schrieb:
              

Fernando Lopez-Lezcano a écrit :
                

A more basic question... what parameters are you using when you start
jackd?
      
              
                  

.jackdrc:

/usr/bin/jackd -R -P70 -t1000 -u -dalsa -r44100 -p64 -n2 -D -Chw:0,1 
-Phw:0,0 -S -i2 -o2
            
                

USB-Devices often perform much better with -n3. Furthermore I would try
to eliminate any flag, that is not absolutely needed such as:

/usr/bin/jackd -R -dhw:1 -dalsa -r44100 -p64 -n3

BTW: this would compute to around 4 ms latency - is that good enough for
you?
          
              

Okay, i had forgot to turn off the system "bips" and esd (it
reduced the number of xruns)
And yep with -n3 i 've got even less xruns and it seems they're
not audible. 
I don't know if the latency is low enough though, i'll test this
evening with my drum pads.

Thanks everyone , it's really great to get so much help so
quickly. 

        
            

Argh, actually i still have audible xruns. And i don't think i'd be
able to play correctly with a higher latency anyway...
Think i'll have to get back to ubuntu studio ... :( 
      
          

Actually after a reboot and some more tweakings, things have been
running without a xrun for an hour now, even with periods/buffer=2.

So finally, i'll stay with planet ccrma ;) 
    
        

You might want to try the 2.6.22.x based kernel for f8 as stated
somewhere else in the thread. In my experience it has better performance
latency-wise. 

What other tweakings did you do?
      

Ok finally, i tried to disable the second core ( i have a dual core
2GHz) and things get really better this way, i can lauch more
applications with 4,8ms latency and no xruns. I had already disabled
cpuspeed but it looks like it was not enough. 
    

That points to a realtime kernel problem with (probably) timekeeping in
two core processors (ie: "time of day" is not perfectly synchronized
between the processors so that if jack migrates from one to another it
appears as there has been too much delay and things don't work as
intended). I have seen this happen in 2.6.23.x + rt14. 

What happens if you boot dual core with the "idle=poll" kernel
parameter? (I assume this happens with 2.6.22.x). 

-- Fernando