Atte André Jensen wrote:
Jack O'Quin wrote:
"Best" depends on your needs and
priorities.
I would accept a kernel that is only used for music. Actually the only
thing I need realtime for is running csound5 in realtime controlled over
midi. I just upgraded from csound4 and performance is better in csound5,
but still I cannot get buffersize lower than 512 (csound-alone-latency:
12ms = too much) without a click every 30 secs or so. csound4 clicked
much more, about every 2 secs, so I'm thinking I'm close, mainly due to
the fact csound5 is able to use the alsa drivers instead of
oss-emulation.
After weeks of patching, re-patching, failed patches, on a Startcom
Distro, I installed FC3 > CCRMA, 2.6.10-0.6.rdt.rhfc3.ccrma &
2.6.10-2.1.ll.rhfc3.ccrma kernels > audio apps. The * -2.1.ll crashed a
couple of times so I stayed with the -.0.6.rtd and had it installed,
upgraded, functional and stable in hours.
With some experimental configuring it is running at .667ms qjackctl,
ardour, hydrogen, on-board AC97 sound with 0 Xruns. Loading jack_rack,
qsynth, rosegarden and Zyaddsubfx through qjackctl hasn't caused an
xrun but I haven't played or recorded with these yet.
For the easiest solution, go with PlanetCCRMA
(Fedora/RedHat) or
AGNULA/DeMuDi (Debian). They've got this stuff all integrated and
readily available for binary download.
I'm not interrested in "easy" but in "best". I'm on
debian/unstable, so
maybe agnula would be possible. I just want to make sure that my current
system is not "infected" with all kinds of agnula stuff. Is it possible
just to get the low-latency kernel and use on an unstable system?
For myself CCRMA was win/win, easy/best.
Just curious if others are getting latency down to .677 +- or am I
fooling myself.
I've yet to get a latency-test program to work.
Athalon oc'd to 3200+ 1G ram 120g hd
qjackctl; Frames/Period=16 Sample Rate=48000 Peroids/Buffer 2
Ardour; jack>latency>32= 0.7msecs
hdparm -m 8 -d 1 -u 1 -c 1 -K 1 /dev/hda/
lspci
00:06.0 Multimedia audio controller: nVidia Corporation nForce2 AC97
Audio Controler (MCP) (rev a1)
01:08.0 Ethernet controller: Intel Corp. 82557/8/9 [Ethernet Pro 100]
(rev 05)
|
from
http://www-106.ibm.com/developerworks/library/l-hw2.html
#"open up" the PCI bus by allowing fairly long bursts for all devices,
increasing performance
setpci -v -d *:* latency_timer=b0
#maximize latency timers for network and audio, allowing them to transmit
#more data per burst, preventing buffer over/underrun conditions
setpci -v -s 00:06.0 latency_timer=ff
setpci -v -s 01:08.0 latency_timer=ff |