11 Feb
2004
11 Feb
'04
7:15 a.m.
hello everyone!
this is not exactly about audio-development, but the DRI+jack thread
encouraged me to ask a few questions about optimizing my setup:
i'm using a rme hdsp (cardbus) + multiface with my 1ghz pIII laptop for
quite a while now, and i'm happy with it.
i would like to use the setup for live-processing but the lowest jackd
framesize i can safely use is 2048. the main issue here seems to be the
nvidia graphics card.
since i switched from kernel 2.4.2x+lowlat+capabilities to 2.6.1 the
cpu-load seems to be noticable higher for ardour, but the latency-issue
did not change.
so my questions are:
*) is anybody successfully using hdsp-cardbus with really low latencies ?
*) how important is the firmware version - which firmware version should
i use for my hdsp card?
*) did anyone else notice increasing cpu load with audio-apps + kernel
2.6.x ?
*) any hints about configuring X11 for linux-audio (besides from DRI
questions ;-) ?
let me know if you think i'm on the wrong list should better alsa-dev.
thanks in advance
regards
d13b
--
Military intelligence is a contradiction in terms.
-- Groucho Marx
11 Feb
11 Feb
1:38 p.m.
d wrote:
hello everyone!
this is not exactly about audio-development, but the DRI+jack thread
encouraged me to ask a few questions about optimizing my setup:
i'm using a rme hdsp (cardbus) + multiface with my 1ghz pIII laptop for
quite a while now, and i'm happy with it.
i would like to use the setup for live-processing but the lowest jackd
framesize i can safely use is 2048. the main issue here seems to be the
nvidia graphics card.
since i switched from kernel 2.4.2x+lowlat+capabilities to 2.6.1 the
cpu-load seems to be noticable higher for ardour, but the latency-issue
did not change.
so my questions are:
*) is anybody successfully using hdsp-cardbus with really low latencies ?
I'm using the same setup as you (hdsp+cardbus+1GHz PIII+nvidia binary
drivers+2.6.1) and can use low latency without any problem. I use jack
with the -R flag as a user thanks to Jack O'Quin's excellent 'realtime'
module. Before that I used 2.4.x-ck and was also quite happy with it.
The only latency killer apps I could identify at the time were
applications accessing ACPI features (notably gnome's batstat applet).
The ACPI situation has been much improved for me in the 2.6.x kernels
(thanks to the CONFIG_ACPI_RELAXED_AML option - I have a toshiba laptop
with a buggy BIOS)
*) how important is the firmware version - which
firmware version should
i use for my hdsp card?
Firmware version shouldn't change anything.
*) did anyone else notice increasing cpu load with
audio-apps + kernel
2.6.x ?
I didn't.
*) any hints about configuring X11 for linux-audio
(besides from DRI
questions ;-) ?
You could try the open source nv driver that's shipped with X. You'll
loose 3rd acceleration, though.
let me know if you think i'm on the wrong list should better alsa-dev.
I don't think for now that this is a driver issue, but can be wrong.
What version of alsa are you using ?
What happens when you run jack with period size < 2048 ? I guess jack
reports xruns, but do they seem to appear regularly or randomly ? In
your description you seem to imply that you do not have this latency
problem when running without X, is it the case ? How low can you get
running jack in text mode ? Generally speaking, can you find any
relation between the xruns and system activity (disk access, incoming
midi stream, GUI interaction, etc...) ?
Thomas
4:08 p.m.
*) is anybody successfully using hdsp-cardbus with
really low latencies ?
many people, including myself, use it down to 64 frames (i have a few
problems at 64 frames under heavy load).
*) how important is the firmware version - which
firmware version should
i use for my hdsp card?
not relevant.
*) did anyone else notice increasing cpu load with
audio-apps + kernel
2.6.x ?
2.6.x is known not to be as good as 2.4+low_lat
*) any hints about configuring X11 for linux-audio
(besides from DRI
questions ;-) ?
not from me, sorry.
--p
12 Feb
12 Feb
7:25 a.m.
thank you thomas an paul!
after a night of configuring an testing i can safely go down to 512
frames. 256 is a bit xruny.
the whole issue seems to be related to my specific laptop. everything
(e.g. 128 frames) works fine on a new centrino laptop with the same
software.
things i noticed (in case someone has a similar setup):
*) i downgraded the bios of my dell i8000 from version A22 to A20 which
brought the biggest improvement. acpi seems to be handled differently in
newer bois versions.
*) using X11 with nv driver instead of nvidia's closed source driver
brings slight improvements. jackd without xserver running also brings
better results (no dramatic changes though)
*) kernel 2.4.22+lowlat+capabilities performs far better for me than
2.6.1+cap+realtime module
*) CONFIG_ACPI_RELAXED_AML does not seem to change the situation on my dell.
regards
d13b
>*) is anybody successfully using hdsp-cardbus with
really low latencies ?
--
Military intelligence is a contradiction in terms.
-- Groucho Marx
7529
days inactive
7530
days old
linux-audio-dev@lists.linuxaudio.org
3 comments
3 participants
participants (3)
-
d -
Paul Davis -
Thomas Charbonnel