On Wed, 15 Feb 2006 20:40:56 +0100 "emanuele ..:: www.rumoridifondo.com ::.." <emanuele(a)rumoridifondo.com> wrote: The -rt patch is not a magic thing to make your computers have doubled cpu strength. I'd suggest a simple test of which the result would interest me.. Play one note, then play two at the same time, then three. And so on. Watch some cpu load program during this (gkrellm, gnome-system-monitor, top, htop, etc. come to my mind). Pretty much independent of the softsynth you use, the cpu load should go up when playing more and more voices. At some point in time your cpu just isn't fast enough anymore. I wonder whether this is what you're seeing. Just to clarify what the -rt patch actually does: Given that your hardware functions and your priorities in the system are setup correctly, you should be able to achieve ridicously low latency settings in jack. Keep in mind though, that this comes at a cost. The CPU overhead raises significantly depending on what software you use with shorter latencies (The cpu has to switch tasks much more often with small buffer sizes). The other advantage of a -rt kernel is that pretty much regardless of what latency setting you use you can almost guarantee that no xrun will ever happen, except for ones induced by faulty jack clients. So even for people without need for supershort latencies or even moderately short latencies (i.e. people who only want to record and do no live input->output processing), the irq priorization setup (and jackd) gives another level of safeness from xruns. On a vanilla kernel an IRQ can disrupt jack operation for pretty long times. Properly setup this cannot happen on a -rt system. About the setup of irq handlers: The chrt program usually comes in a package called schedutils. Look out for that or build it from source. Flo -- Palimm Palimm! http://tapas.affenbande.org