Re: [proaudio] trouble emerging museseq

[ Thread Index | Date Index | More lists.tuxfamily.org/proaudio Archives ]


RT-kernel are very sensitive to bad hardware configuration. IRQ can be
a big source of problem. You must be sure at you don't share some IRQ.
# cat /proc/interrupts
           CPU0
  0:   68976887  IO-APIC-edge   [........N/  0]  pit
  1:      29813  IO-APIC-edge   [........./  1]  i8042
  7:          0  IO-APIC-edge   [........./  0]  parport0
  8:          2  IO-APIC-edge   [........./  0]  rtc
 14:     650128  IO-APIC-edge   [........./  0]  ide0
 15:     173813  IO-APIC-edge   [........./  0]  ide1
169:     504475  IO-APIC-level  [........./  0]  acpi, ohci_hcd:usb1
177:          0  IO-APIC-level  [........./  0]  Ensoniq AudioPCI
185:    4003455  IO-APIC-level  [........./  0]  EMU10K1
193:          0  IO-APIC-level  [........./  0]  ohci_hcd:usb2
201:         16  IO-APIC-level  [........./  0]  ohci_hcd:usb3
209:    7546427  IO-APIC-level  [........./  0]  ehci_hcd:usb4
217:   20331579  IO-APIC-level  [........./  0]  ohci1394, eth0
225:          6  IO-APIC-level  [........./  0]  bttv0
NMI:          0
LOC:    6252476
ERR:          0
MIS:          0

As you can see, I have shared irq on irq 169 and 217. It is not perfect but work fine.
The most important is the sound and video.
I have an Nvidia car and don't use the 3D nvidia driver because it use one more IRQ that is shared with the Ensonic.
I use APIC. Some people don't like it because it have been repported many problems with windows when they shifted to APIC. I never have problem with it and linux. If you are using PIC, it is the same, shared IRQ is not good with a RT kernel. I believe at it will be the same problematic with a PAM-Rlimit system. 
The device sharing the IRQ with the sound device will get the same priority as the sound device, and it can cause many trouble for the kernel, even system crash or freezing hardware.

But in your case, I think at 512 MB RAM is too little to do serious audio work. It is what was in my box, and I added 1GB and many problems as intermitant noises was gone. I recommand for an audio workstation to have the same hardware as a graphic station, high tech motherboard, huge ram memory, but with simpler graphic card (less noise footprint), better sound card (it is what we want) and better hdd system (a must for multitrack recording).

After adding this ram, I can run jack for hours or even days without problems. I would say at with the right hardware, jack will work fine and at it is not a problem which kernel you are using. The performances can differ with the kernel, but not the sound quality.

You must be sure at jack is using tmpfs (USE jack-tmpfs). Check in fstab at you have:
none	/dev/shm		tmpfs	nodev,nosuid,noexec	0 0
Otherwise, jackd will use the hdd to do its work, and it can be a hudge amount of data. tmpfs will use RAM  but the amount of data will remain the same, it will work faster and with less noise.

Dominique

Le Wed, 17 May 2006 22:11:14 +0200,
SteelRage <keywiz@xxxxxxxxxxx> a écrit :

> Robert Persson wrote:On Wednesday 17 May 2006 05:12 Thomas Kuther was like:
>   hmmm, which just seconds the comment by some (former) LADs on a
> webblog somewhere that blamed rosegarden devs for including new features
> instead of concentrating on the core functionality first. Here it even
> freezes X when i try to start it.  
>     The most frustrating linux audio problem I have is actually with jack, 
> although rosegarden is very likely to trigger it, of course. If the load on 
> the system gets very heavy strange distortions and other artifacts sometimes 
> creep into the sound, and they don't go away even when things have settled 
> down a bit (at least I think that's what's happening). Whatever it is, it 
> doesn't correct itself until I restart jack.
> 
>    Just a noob question (sorry if you already did that)...
> Are you using a RT-kernel? I had that problems too without it.
> Now (I installed it yesterday) I can run rosegarden with VSTi/dssi synths/qsynth with a 128 MB soundfont (didn't try other softsynths) at 2.5ms of latency WHILE COMPILING (!!!!), without all that big problems you wrote about.
> It's been a DRAMATIC improvement.
> Ah, there were also a session of kvirc, kopete, thunderbird and firefox with about 6-8 tabs opened.
> 
> My system is an athlon XP 2000+ with 512 MB of RAM and an Echo-Digital MIA.
> It can run also at 1.5 ms of latency (without all that stuff opened), but I preferred to trade a bit of latency with more reliabiliy :o
> 
> Paolo



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/