23 Jan
2015
23 Jan
'15
11:57 a.m.
I've get a new (used) mobo, and played with the kernel version to use.
So far I've tested all from 3.19.0-rc5 down to 3.4.104
Those, were rt-patches available I test as rt-kernel.
Best results, I only get with the 3.4 series, all later kernels
introduce Xruns here, sporadic, unrelated to the dsp load, even, when
just jack is running.
I build then all with the same configuration, play a bit with
configurations deselect hyper-threading on the latest kernels, use
threadedirqs, all to no avail. Only 3.4.xxx-rt kernel run absolute smooth.
So, which kernels with witch configuration do you all use?
System: Host: box Kernel: 3.4.105-rt129 x86_64 (64 bit gcc: 4.9.2)
Desktop: Cinnamon 2.2.16 (Gtk 3.14.5) dm: lightdm
Distro: siduction 14.1.0 Indian Summer - cinnamon -
(201411230307)
Machine: Mobo: Intel model: DG965MQ v: AAD37419-302
Bios: Intel v: MQ96510J.86A.1250.2006.1005.1532 date: 10/05/2006
CPU: Quad core Intel Core2 Quad Q6600 (-MCP-) cache: 4096 KB
flags: (lm nx sse sse2 sse3 ssse3 vmx) bmips: 19200
clock speeds: min/max: 1596/2394 MHz 1: 2394 MHz 2: 2394 MHz
3: 2394 MHz 4: 2394 MHz
Graphics: Card: NVIDIA GT216 [GeForce GT 220]
bus-ID: 01:00.0 chip-ID: 10de:0a20
Display Server: X.Org 1.16.1.901 drivers: nouveau (unloaded:
fbdev,vesa)
Resolution: 1600x900(a)59.98hz
GLX Renderer: Gallium 0.4 on NVA5
GLX Version: 3.0 Mesa 10.3.2 Direct Rendering: Yes
Audio: Card-1 NVIDIA GT216 HDMI Audio Controller
driver: snd_hda_intel bus-ID: 01:00.1 chip-ID: 10de:0be2
Card-2 Creative Labs SB Audigy
driver: snd_emu10k1 port: 1000 bus-ID: 07:00.0 chip-ID:
1102:0004
Sound: Advanced Linux Sound Architecture v: 1.0.25
Drives: HDD Total Size: 164.7GB (26.4% used)
ID-1: /dev/sda model: HDT722516DLA380 size: 164.7GB serial:
VDK71CTCEBBN5X
23 Jan
23 Jan
12:13 p.m.
On 01/23/2015 11:57 AM, Hermann Meyer wrote:
I've get a new (used) mobo, and played with the
kernel version to use.
So far I've tested all from 3.19.0-rc5 down to 3.4.104
Those, were rt-patches available I test as rt-kernel.
Best results, I only get with the 3.4 series, all later kernels
introduce Xruns here, sporadic, unrelated to the dsp load, even, when
just jack is running.
I build then all with the same configuration, play a bit with
configurations deselect hyper-threading on the latest kernels, use
threadedirqs, all to no avail. Only 3.4.xxx-rt kernel run absolute smooth.
So, which kernels with witch configuration do you all use?
System: Host: box Kernel: 3.4.105-rt129 x86_64 (64 bit gcc: 4.9.2)
Desktop: Cinnamon 2.2.16 (Gtk 3.14.5) dm: lightdm
Distro: siduction 14.1.0 Indian Summer - cinnamon -
(201411230307)
Machine: Mobo: Intel model: DG965MQ v: AAD37419-302
Bios: Intel v: MQ96510J.86A.1250.2006.1005.1532 date: 10/05/2006
CPU: Quad core Intel Core2 Quad Q6600 (-MCP-) cache: 4096 KB
flags: (lm nx sse sse2 sse3 ssse3 vmx) bmips: 19200
clock speeds: min/max: 1596/2394 MHz 1: 2394 MHz 2: 2394 MHz
3: 2394 MHz 4: 2394 MHz
Graphics: Card: NVIDIA GT216 [GeForce GT 220]
bus-ID: 01:00.0 chip-ID: 10de:0a20
Display Server: X.Org 1.16.1.901 drivers: nouveau (unloaded:
fbdev,vesa)
Resolution: 1600x900(a)59.98hz
GLX Renderer: Gallium 0.4 on NVA5
GLX Version: 3.0 Mesa 10.3.2 Direct Rendering: Yes
Audio: Card-1 NVIDIA GT216 HDMI Audio Controller
driver: snd_hda_intel bus-ID: 01:00.1 chip-ID: 10de:0be2
Card-2 Creative Labs SB Audigy
driver: snd_emu10k1 port: 1000 bus-ID: 07:00.0 chip-ID:
1102:0004
Sound: Advanced Linux Sound Architecture v: 1.0.25
Drives: HDD Total Size: 164.7GB (26.4% used)
ID-1: /dev/sda model: HDT722516DLA380 size: 164.7GB serial:
VDK71CTCEBBN5X
Hi Hermann,
On my Quad Core Core2 machine I'm still on 3.2 RT. Works like a charm, I
only get xruns when using very low latency settings like -p8 -n3 -r48000.
Jeremy
12:59 p.m.
Am 23.01.2015 um 12:13 schrieb Jeremy Jongepier:
On 01/23/2015 11:57 AM, Hermann Meyer wrote:
Hi Jeremy
Well, I guess the 3.2 RT will work here as well, just kernels > 3.4
started to behave bad.
The main reason I would use later kernels, special >3.18 is, that they
have a big improvement in the nouveau driver. Lately the 3.19.0-rc5
detect my card correct and handle the GPU mem correct.
hermann
I've get a new (used) mobo, and played with
the kernel version to use.
So far I've tested all from 3.19.0-rc5 down to 3.4.104
Those, were rt-patches available I test as rt-kernel.
Best results, I only get with the 3.4 series, all later kernels
introduce Xruns here, sporadic, unrelated to the dsp load, even, when
just jack is running.
I build then all with the same configuration, play a bit with
configurations deselect hyper-threading on the latest kernels, use
threadedirqs, all to no avail. Only 3.4.xxx-rt kernel run absolute smooth.
So, which kernels with witch configuration do you all use?
Hi Hermann,
On my Quad Core Core2 machine I'm still on 3.2 RT. Works like a charm, I
only get xruns when using very low latency settings like -p8 -n3 -r48000.
Jeremy
1:43 p.m.
On 01/23/2015 12:59 PM, Hermann Meyer wrote:
Hi Jeremy
Well, I guess the 3.2 RT will work here as well, just kernels > 3.4
started to behave bad.
The main reason I would use later kernels, special >3.18 is, that they
have a big improvement in the nouveau driver. Lately the 3.19.0-rc5
detect my card correct and handle the GPU mem correct.
hermann
Hello Hermann,
Ah nouveau, as I'm using a fairly old nvidia card I'm not having any
issues with nouveau. But I can always try a newer kernel on my setup if
you'd like. Are you using the onboard audio or external gear?
Jeremy
3:58 p.m.
Am 23.01.2015 um 13:43 schrieb Jeremy Jongepier:
On 01/23/2015 12:59 PM, Hermann Meyer wrote:
Well, I'm not really having issues with nouveau, just those error
messages in dmesg
[ 21.627203] [drm] nouveau 0000:01:00.0: illegal object class: 0xa0b5
[ 21.627208] [drm] nouveau 0000:01:00.0: Error creating object: -22
(3/0x0000a0b5)
[ 21.627212] [drm] nouveau 0000:01:00.0: illegal object class: 0x90b8
[ 21.627215] [drm] nouveau 0000:01:00.0: Error creating object: -22
(3/0x000590b8)
[ 21.627218] [drm] nouveau 0000:01:00.0: illegal object class: 0x90b5
[ 21.627221] [drm] nouveau 0000:01:00.0: Error creating object: -22
(3/0x000490b5)
Which didn't appear with newer Kernels.
Also, I'm just curious about what going on in the newer Kernels, as,
when I set up this MOBO, it's a work station, no wireless, I've to move
down the road to older kernels, till I end up at 3.4.104, as the first
one which works smooth. I use a PCI Creative Labs SB Audigy, onboard
sound is disabled, but the Nvidia Card provide a Audio output as well.
I've connected my Monitor via HDMI cable.
With kernel 3.4.104 I can use pulseaudio(sink) in jack down to 128
frames/period without Xruns,
but with younger kernels, Xruns appear randomness, regardless of use
even if pulse is suspend and just jack runs with a dsp load of 0.33%.
I've set CPU freq to performance (in kernel config), I ensure that the
soundcard own his own IRQ, boost the priory of it, disable
hyper-threading, disable irqbalancing, but nothing helps.
Non of this tweaks is needed for the 3.4.104/5 - rt kernels. They work
out of the box, even when my soundcard share a IRQ with a USB hub, no
problem at all.
However, as I said, I'm just curios about what others here use, I'm
perfectly fine with 3.4.105-rt129, but still I would experiment more
with recent kernels, and I like to find out what cause the Xruns, as I
like to use a newer nouveau driver, .
regards
hermann
Hi Jeremy
Well, I guess the 3.2 RT will work here as well, just kernels > 3.4
started to behave bad.
The main reason I would use later kernels, special >3.18 is, that they
have a big improvement in the nouveau driver. Lately the 3.19.0-rc5
detect my card correct and handle the GPU mem correct.
hermann
Hello Hermann,
Ah nouveau, as I'm using a fairly old nvidia card I'm not having any
issues with nouveau. But I can always try a newer kernel on my setup if
you'd like. Are you using the onboard audio or external gear?
Jeremy
10:58 p.m.
On Fri, 23 Jan 2015, Hermann Meyer wrote:
However, as I said, I'm just curios about what
others here use, I'm perfectly
fine with 3.4.105-rt129, but still I would experiment more with recent
kernels, and I like to find out what cause the Xruns, as I like to use a
newer nouveau driver, .
$ uname -a
Linux music 3.13.0-43-lowlatency #72-Ubuntu SMP PREEMPT Mon Dec 8 20:00:12
UTC 2014 x86_64 x86_64 x86_64 GNU/Linux
ICE1712 PCI audio card. Runs Jack at 48k 16/2 with no xruns, Runs Pulse
jacksink at least down to 32/2 as well. (I haven't tried it lower) Intel
on board video. I also have an old ensoniq Audio card (PCI) that I use
only for MIDI.
On my old P4 MB I did have trouble with xruns if the ensoniq was on a
higher IRQ than the ICE1712, but it doesn't seem to matter on this new MB
(i5 based) where the ICE is irq17 and the ens1370 is irq 19.
Maybe I have just been lucky, but I have never felt the need for an RT
kernel yet.
--
Len Ovens
www.ovenwerks.net
12:45 p.m.
On 23/01/15 at 11:57am, Hermann Meyer wrote:
I've get a new (used) mobo, and played with the
kernel version to use. So
far I've tested all from 3.19.0-rc5 down to 3.4.104
Those, were rt-patches available I test as rt-kernel.
Best results, I only get with the 3.4 series, all later kernels introduce
Xruns here, sporadic, unrelated to the dsp load, even, when just jack is
running.
I build then all with the same configuration, play a bit with configurations
deselect hyper-threading on the latest kernels, use threadedirqs, all to no
avail. Only 3.4.xxx-rt kernel run absolute smooth.
So, which kernels with witch configuration do you all use?
System: Host: box Kernel: 3.4.105-rt129 x86_64 (64 bit gcc: 4.9.2)
Desktop: Cinnamon 2.2.16 (Gtk 3.14.5) dm: lightdm
Distro: siduction 14.1.0 Indian Summer - cinnamon -
(201411230307)
Machine: Mobo: Intel model: DG965MQ v: AAD37419-302
Bios: Intel v: MQ96510J.86A.1250.2006.1005.1532 date: 10/05/2006
CPU: Quad core Intel Core2 Quad Q6600 (-MCP-) cache: 4096 KB
flags: (lm nx sse sse2 sse3 ssse3 vmx) bmips: 19200
clock speeds: min/max: 1596/2394 MHz 1: 2394 MHz 2: 2394 MHz
3: 2394 MHz 4: 2394 MHz
Graphics: Card: NVIDIA GT216 [GeForce GT 220]
bus-ID: 01:00.0 chip-ID: 10de:0a20
Display Server: X.Org 1.16.1.901 drivers: nouveau (unloaded:
fbdev,vesa)
Resolution: 1600x900(a)59.98hz
GLX Renderer: Gallium 0.4 on NVA5
GLX Version: 3.0 Mesa 10.3.2 Direct Rendering: Yes
Audio: Card-1 NVIDIA GT216 HDMI Audio Controller
driver: snd_hda_intel bus-ID: 01:00.1 chip-ID: 10de:0be2
Card-2 Creative Labs SB Audigy
driver: snd_emu10k1 port: 1000 bus-ID: 07:00.0 chip-ID: 1102:0004
Sound: Advanced Linux Sound Architecture v: 1.0.25
Drives: HDD Total Size: 164.7GB (26.4% used)
ID-1: /dev/sda model: HDT722516DLA380 size: 164.7GB serial:
VDK71CTCEBBN5X
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-user
debian based, why don't you try one of this
https://packages.debian.org/search?searchon=all&keywords=PREEMPT_RT
then you can also compile from vanilla with 'make oldconfig' and adjust relevant
settings
according to your needs.
I recently compiled longterm 3.14.29 with realtime patch
https://www.kernel.org/pub/linux/kernel/projects/rt/3.14/
works like a charm
--
«My mama said
to get things done
You'd better not mess
with Major Tom»
D.Bowie
12:53 p.m.
Am 23.01.2015 um 12:45 schrieb Raffaele Morelli:
debian based, why don't you try one of this
https://packages.debian.org/search?searchon=all&keywords=PREEMPT_RT
then you can also compile from vanilla with 'make oldconfig' and adjust relevant
settings
according to your needs.
I used to build my own kernels from vanilla since 2.16 or earlier,
didn't remember really :-[
I recently compiled longterm 3.14.29 with realtime
patch
https://www.kernel.org/pub/linux/kernel/projects/rt/3.14/
works like a charm
I've tried 3.14.25-rt22 and 3.14.28-rt25, both gime Xruns.
Are you using Quad CoreCPU? As, with my previous Mobo with Single Core
CPU, I've used the 3.14.25 as well, and it works like a charm there as
well.
1:09 p.m.
On 23/01/15 at 12:53pm, Hermann Meyer wrote:
Am 23.01.2015 um 12:45 schrieb Raffaele Morelli:
I am using a relatively old dual core amd athlon X2 (I'll send you details in the
afternoon as I am
actually on another machine).
You got Xruns? under which conditions?
--
«My mama said
to get things done
You'd better not mess
with Major Tom»
D.Bowie
debian based, why don't you try one of this
https://packages.debian.org/search?searchon=all&keywords=PREEMPT_RT
then you can also compile from vanilla with 'make oldconfig' and adjust relevant
settings
according to your needs.
I used to build my own kernels from vanilla since 2.16 or earlier, didn't
remember really :-[
I recently compiled longterm 3.14.29 with realtime
patch
https://www.kernel.org/pub/linux/kernel/projects/rt/3.14/
works like a charm
I've tried 3.14.25-rt22 and 3.14.28-rt25, both gime Xruns.
Are you using Quad CoreCPU? As, with my previous Mobo with Single Core CPU,
I've used the 3.14.25 as well, and it works like a charm there as well.
1:46 p.m.
On 23/01/15 at 12:53pm, Hermann Meyer wrote:
Am 23.01.2015 um 12:45 schrieb Raffaele Morelli:
model name : AMD Athlon(tm) 64 X2 Dual Core Processor 5200+
cpu MHz : 2700.311
cache size : 512 KB
I gave you wrong kernel release, I am also using 3.14.25
My config-3.14.25-rt22 config
http://fpaste.org/173515/
--
« Nunc est bibendum, nunc pede libero pulsanda tellus »
debian based, why don't you try one of this
https://packages.debian.org/search?searchon=all&keywords=PREEMPT_RT
then you can also compile from vanilla with 'make oldconfig' and adjust relevant
settings
according to your needs.
I used to build my own kernels from vanilla since 2.16 or earlier, didn't
remember really :-[
I recently compiled longterm 3.14.29 with realtime
patch
https://www.kernel.org/pub/linux/kernel/projects/rt/3.14/
works like a charm
I've tried 3.14.25-rt22 and 3.14.28-rt25, both gime Xruns.
Are you using Quad CoreCPU? As, with my previous Mobo with Single Core CPU,
I've used the 3.14.25 as well, and it works like a charm there as well. 
3:34 p.m.
Hi Hermann,
on my machine issues started with kernels > 3.8.13-rt14. They locked my
machine. After a while newer versions of linux-rt worked again, but I
didn't used them much for audio work, so I don't know how good they work.
E.g. 3.10.61-rt65 does work (available as linux-rt-lts from Arch Audio),
3.14.28-rt25 does work (available as linux-rt from Arch Audio), but at
least CPU frequency scaling is broken for 3.14.28-rt25 and it was the same
for 3.14.25-rt22. It's impossible to change the governor, so if you for
some reason would like to use ondemand instead of performance, it doesn't
work. IOW my experiences were good for IIRC <= 3.8.13-rt14 and nowadays I
would test 3.10.61-rt65 and/or 3.14.28-rt25 for audio work. I only would
downgrade to 3.8.13-rt14 if I would experience issues. At the moment I'm
using 3.10.61-rt65 for averaged desktop usage, because it can be used with
frequency scaling ondemand and I can build vbox modules for the last
really stable version of vbox. IIRC >= 3.16 only build modules for broken
versions of vbox, IOW for vbox > 4.3.12, but this version is the last
version with completely good USB support. IOW 3.18.2-2-ARCH is useless for
my needs when I want to run vbox, but it works for anything else. However,
even with threadirqs I never used it for audio work either. The issues
with kernels > 3.8.13-rt14 happened for AMD machines, I wasn't the only
user experiencing issues. Mobo ASUS M2A-VM HDMI, CPU Athlon dual-core
BE-2350 (2.1 GHz), RAM 4 GiB, audio card RME HDSPe AIO (can't be used
without getting xruns), 2 cards for MIDI are TerraTec EWX 24/96.
Resume: IIRC <= 3.8.13-rt14 were ok (on Ubuntu and Debian installs too)
and 3.10.61-rt65 likely is ok too, but not really tested on my machine.
Regards,
Ralf
3:59 p.m.
Am 23.01.2015 um 13:43 schrieb Raffaele Morelli:
On 23/01/15 at 12:53pm, Hermann Meyer wrote:
Thanks, I will compare and try some of your settings, to see if it helps.
regards
hermann
Am 23.01.2015 um 12:45 schrieb Raffaele Morelli:
model
name : AMD Athlon(tm) 64 X2 Dual Core Processor 5200+
cpu MHz : 2700.311
cache size : 512 KB
I gave you wrong kernel release, I am also using 3.14.25
My config-3.14.25-rt22 config
http://fpaste.org/173515/
debian based, why don't you try one of this
https://packages.debian.org/search?searchon=all&keywords=PREEMPT_RT
then you can also compile from vanilla with 'make oldconfig' and adjust relevant
settings
according to your needs.
I used to build my own kernels from vanilla since 2.16 or
earlier, didn't
remember really :-[
I recently compiled longterm 3.14.29 with
realtime patch
https://www.kernel.org/pub/linux/kernel/projects/rt/3.14/
works like a charm
I've tried 3.14.25-rt22 and 3.14.28-rt25, both gime Xruns.
Are you using Quad CoreCPU? As, with my previous Mobo with Single Core CPU,
I've used the 3.14.25 as well, and it works like a charm there as well.
6:15 p.m.
On Fri, 23 Jan 2015 12:45:39 +0100, Raffaele Morelli wrote:
I recently compiled longterm 3.14.29 with realtime
patch
https://www.kernel.org/pub/linux/kernel/projects/rt/3.14/
works like a charm
Assumed you don't need CPU frequency scaling "performance", can you
switch to ondemand?
6:32 p.m.
Am 23.01.2015 um 18:15 schrieb Ralf Mardorf:
On Fri, 23 Jan 2015 12:45:39 +0100, Raffaele Morelli
wrote:
According to Raffaeles .config file, CPU frequency scaling is
completely disabled.
One of the difference to mine config, were I've set frequency scaling to
performance on default.
I recently compiled longterm 3.14.29 with
realtime patch
https://www.kernel.org/pub/linux/kernel/projects/rt/3.14/
works like a charm
Assumed you don't need CPU frequency scaling
"performance", can you
switch to ondemand?
_______________________________________________ 
6:36 p.m.
CPU scaling is broken in "recent" -rt; particularly on older CPU's
(without the intel pstate stuff). I've also built -rt with the
performance governer on as default, and some code in the kernel
dis-allows any changing of CPU frequency.
There's a thread over on -rt users, but it wasn't resolved yet: hence
building with performance. Apart from minor lower battery life, and
some extra heat / fan-noise, there's no dis-advantage.
http://www.spinics.net/lists/linux-rt-users/msg12170.html
HTH, -Harry
On Fri, Jan 23, 2015 at 5:32 PM, Hermann Meyer <brummer-(a)web.de> wrote:
Am 23.01.2015 um 18:15 schrieb Ralf Mardorf:
--
http://www.openavproductions.com
On Fri, 23 Jan 2015 12:45:39 +0100, Raffaele Morelli wrote:
According to Raffaeles .config file, CPU frequency scaling is completely
disabled.
One of the difference to mine config, were I've set frequency scaling to
performance on default.
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-user
I recently compiled longterm 3.14.29 with realtime patch
https://www.kernel.org/pub/linux/kernel/projects/rt/3.14/
works like a charm
Assumed you don't need CPU frequency scaling "performance", can you
switch to ondemand?
_______________________________________________
7:12 p.m.
On Fri, 23 Jan 2015 18:32:07 +0100, Hermann Meyer wrote:
According to Raffaeles .config file, CPU frequency
scaling is
completely disabled.
One of the difference to mine config, were I've set frequency scaling
to performance on default.
On Fri, 23 Jan 2015 17:36:05 +0000, Harry van Haaren wrote:
CPU scaling is broken in "recent" -rt;
particularly on older CPU's
(without the intel pstate stuff). I've also built -rt with the
performance governer on as default, and some code in the kernel
dis-allows any changing of CPU frequency.
There's a thread over on -rt users, but it wasn't resolved yet: hence
building with performance. Apart from minor lower battery life, and
some extra heat / fan-noise, there's no dis-advantage.
http://www.spinics.net/lists/linux-rt-users/msg12170.html
I just wanted to point out, that the CPU frequency scaling governor
might be useful for some computer usage. For audio production
"performance" is all we need.
3.14-rts suffer from this issue.
3.10-rts seems to be ok.
While Herman gets best results with 3.4 series, I still had good luck
with 3.8.13-rt14. I don't want to mount all my old installs I still
kept, but IIRC ...
[rocketmouse@archlinux ~]$ ls -hAl /mnt/debi386/boot/vm*3.8*
-rw-r--r-- 1 root root 2.8M Mar 21
2014 /mnt/debi386/boot/vmlinuz-3.8.13.14-rt30-pae-rocketmouse-1
-rw-r--r-- 1 root root 2.8M Jan 27
2014 /mnt/debi386/boot/vmlinuz-3.8.13-rt14-pae-rocketmouse-2
... there were other builds ok too for 3.8 series, at least
3.8.13.14-rt30 ;). But as mentioned before, I'm not using an aged Intel
machine. I'm using an aged AMD machine. Don't worry about the PAE, I
used them with 32-bit and 64-bit architecture.
I don't remember if I used the aged onboard ATI, as I do now, or my
aged PCIe NVIDIA. Usually both aged graphics work ok with the FLOSS
drivers, when using Xfce4, Mate, JWM and openbox or similar, IIRC e.g.
Razor-Qt. They fail when using Cinnamon. I didn't really use Mate, but
Xfce4 a lot and nowadays I prefer JWM and openbox.
7:46 p.m.
On Fri, 23 Jan 2015 19:12:32 +0100
Ralf Mardorf <ralf.mardorf(a)alice-dsl.net> wrote:
On Fri, 23 Jan 2015 18:32:07 +0100, Hermann Meyer
wrote:
This is all very depressing!
I may need to get a laptop in the near future (for the first time) specifically
for audio, and I fear the possibility of a rather expensive mistake... or two :(
--
Will J Godfrey
http://www.musically.me.uk
Say you have a poem and I have a tune.
Exchange them and we can both have a poem, a tune, and a song.
According to Raffaeles .config file, CPU
frequency scaling is
completely disabled.
One of the difference to mine config, were I've set frequency scaling
to performance on default.
On Fri, 23 Jan 2015 17:36:05 +0000, Harry van Haaren wrote:
CPU scaling is broken in "recent" -rt;
particularly on older CPU's
(without the intel pstate stuff). I've also built -rt with the
performance governer on as default, and some code in the kernel
dis-allows any changing of CPU frequency.
There's a thread over on -rt users, but it wasn't resolved yet: hence
building with performance. Apart from minor lower battery life, and
some extra heat / fan-noise, there's no dis-advantage.
http://www.spinics.net/lists/linux-rt-users/msg12170.html
I just wanted to point out, that the CPU frequency scaling governor
might be useful for some computer usage. For audio production
"performance" is all we need.
3.14-rts suffer from this issue.
3.10-rts seems to be ok.
While Herman gets best results with 3.4 series, I still had good luck
with 3.8.13-rt14. I don't want to mount all my old installs I still
kept, but IIRC ...
[rocketmouse@archlinux ~]$ ls -hAl /mnt/debi386/boot/vm*3.8*
-rw-r--r-- 1 root root 2.8M Mar 21
2014 /mnt/debi386/boot/vmlinuz-3.8.13.14-rt30-pae-rocketmouse-1
-rw-r--r-- 1 root root 2.8M Jan 27
2014 /mnt/debi386/boot/vmlinuz-3.8.13-rt14-pae-rocketmouse-2
... there were other builds ok too for 3.8 series, at least
3.8.13.14-rt30 ;). But as mentioned before, I'm not using an aged Intel
machine. I'm using an aged AMD machine. Don't worry about the PAE, I
used them with 32-bit and 64-bit architecture.
I don't remember if I used the aged onboard ATI, as I do now, or my
aged PCIe NVIDIA. Usually both aged graphics work ok with the FLOSS
drivers, when using Xfce4, Mate, JWM and openbox or similar, IIRC e.g.
Razor-Qt. They fail when using Cinnamon. I didn't really use Mate, but
Xfce4 a lot and nowadays I prefer JWM and openbox.
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-user
8:04 p.m.
New subject: OT: Which kernel do you use?
On Fri, 23 Jan 2015 18:46:24 +0000, Will Godfrey wrote:
I fear the possibility of a rather expensive mistake
Your fear shouldn't be related to this thread. When using Linux, we
always risk to make expensive mistakes regarding to
software-hardware-compatibility. OTOH regarding to ethics, the risk is
very low :p, just avoid installing Ubuntu, that by default installs
spyware.
Regarding to software-hardware-compatibility I made much better
experiences with hardware from vendors, who pointed out, that they
don't care about Linux, than with hardware from vendors who claimed
that their hardware does work with Linux or with hardware recommended
by Linux communities.
24 Jan
24 Jan
9:42 a.m.
2015-01-23 18:15 GMT+01:00 Ralf Mardorf <ralf.mardorf(a)alice-dsl.net>et>:
On Fri, 23 Jan 2015 12:45:39 +0100, Raffaele Morelli
wrote:
I totally removed CPU frequecy scaling from the kernel
I recently compiled longterm 3.14.29 with
realtime patch
https://www.kernel.org/pub/linux/kernel/projects/rt/3.14/
works like a charm
Assumed you don't need CPU frequency scaling "performance", can you
switch to ondemand?
23 Jan
23 Jan
7:29 p.m.
On 01/23/2015 11:57 AM, Hermann Meyer wrote:
So, which kernels with witch configuration do you all
use?
atte@skagen:~$ uname -r
3.2.0-4-rt-686-pae
Works very well, crunchbang (=debian wheezy)
--
Atte
http://atte.dk http://a773.dk
7:32 p.m.
On Fri, 23 Jan 2015 11:57:44 +0100, Hermann Meyer wrote:
So far I've tested all from 3.19.0-rc5 down to
3.4.104
Are yo using a localmodconfig like config? When I build kernels myself
I'm copying a default distro's config and run oldconfig and if wanted, I
add a few settings manually before running oldconfig. Building a kernel
that way, on my aged machine takes around 90 minutes. Since a while
back I usually install Arch kernels build by Joakim, instead of building
them on my own.
7:45 p.m.
Am 23.01.2015 um 19:31 schrieb Ralf Mardorf:
On Fri, 23 Jan 2015 11:57:44 +0100, Hermann Meyer
wrote:
Yes, using oldconfig, and afterwards menuconfig or xconfig, if I need to
tweak some more.
With this Quad core it just takes less then half a hour, running in
background.
Well, on the 3.4.xxx-rt kernels I could even build a kernel in
background, and play with guitarix for the time being, at 128
frames/period without any Xrun. On later ones, I can't open a
text-editor without receive a Xrun.
So far I've tested all from 3.19.0-rc5 down
to 3.4.104
Are yo using a localmodconfig like config? When I build kernels myself
I'm copying a default distro's config and run oldconfig and if wanted, I
add a few settings manually before running oldconfig. Building a kernel
that way, on my aged machine takes around 90 minutes. Since a while
back I usually install Arch kernels build by Joakim, instead of building
them on my own.
9:06 p.m.
On 23/01/15 18:45, Hermann Meyer wrote:
Am 23.01.2015 um 19:31 schrieb Ralf Mardorf:
Running stock debian kernel rebuilt with CONFIG_PREEMPT=y (3.18.0 from
experimental at the time), i.e. no rt patch.
Yoshimi through guitarix at 64/3, no xruns, bionc client running in
background.
YMMV
rob
On Fri, 23 Jan 2015 11:57:44 +0100, Hermann Meyer
wrote:
Yes, using oldconfig, and afterwards menuconfig or xconfig, if I need to
tweak some more.
With this Quad core it just takes less then half a hour, running in
background.
Well, on the 3.4.xxx-rt kernels I could even build a kernel in
background, and play with guitarix for the time being, at 128
frames/period without any Xrun. On later ones, I can't open a
text-editor without receive a Xrun.
_______________________________________________ So far I've tested all from 3.19.0-rc5 down
to 3.4.104
Are yo using a localmodconfig like config? When I build kernels myself
I'm copying a default distro's config and run oldconfig and if wanted, I
add a few settings manually before running oldconfig. Building a kernel
that way, on my aged machine takes around 90 minutes. Since a while
back I usually install Arch kernels build by Joakim, instead of building
them on my own.
11:29 p.m.
On 01/23/2015 05:57 AM, Hermann Meyer wrote:
So, which kernels with witch configuration do you all
use?
$ uname -a
Linux chopin 3.17.6-200.fc20.x86_64 #1 SMP Mon Dec 8 15:21:05 UTC 2014
x86_64 x86_64 x86_64 GNU/Linux
Run jack at 64/2 with the ice1712 (delta1010 or 2496) both with
pulseaudio and without. Sporadic xruns when switching workspaces or with
pulseaudio clients entering and exiting quickly but during actual
recording and playback: rock solid.
Running:
- Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz
- 32GB RAM
- 4x 120GB SSD (hardware RAID5)
- 2x 120GB SSD (software RAID0)
- Intel Graphics and Onboard sound (sometimes I give that to pulseaudio)
24 Jan
24 Jan
3:41 a.m.
Am 23.01.2015 um 23:29 schrieb David Santamauro:
On 01/23/2015 05:57 AM, Hermann Meyer wrote:
Maybe I need to try my Onboard Intel Graphic when I move to newer
Kernels, to find out if it is the nouveau driver which case the trouble.
To bad that I start to like the Cinnamon 3D desktop. :-)
regards
hermann
So, which kernels with witch configuration do you
all use?
$ uname -a
Linux chopin 3.17.6-200.fc20.x86_64 #1 SMP Mon Dec 8 15:21:05 UTC 2014
x86_64 x86_64 x86_64 GNU/Linux
Run jack at 64/2 with the ice1712 (delta1010 or 2496) both with
pulseaudio and without. Sporadic xruns when switching workspaces or
with pulseaudio clients entering and exiting quickly but during actual
recording and playback: rock solid.
Running:
- Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz
- 32GB RAM
- 4x 120GB SSD (hardware RAID5)
- 2x 120GB SSD (software RAID0)
- Intel Graphics and Onboard sound (sometimes I give that to
pulseaudio)
7:53 a.m.
On 01/23/2015 04:41 PM, Hermann Meyer wrote:
Am 23.01.2015 um 23:29 schrieb David Santamauro:
I wouldn't be surprised if modern Intel onboard graphics can handle the
Cinnamon 3D desktop. The Intel HD4600 in my laptop handles video stuff
better than the old Intel that was in my previous laptop.
Of course, I'm an old dinosaur. No desktop environment should require
hardware acceleration to work. That way lies the Land of Windows Bloat! ;)
--
David W. Jones
gnome(a)hawaii.rr.com
authenticity, honesty, community
http://dancingtreefrog.com
On 01/23/2015 05:57 AM, Hermann Meyer wrote:
Maybe I need to try my Onboard Intel Graphic when I move to newer
Kernels, to find out if it is the nouveau driver which case the trouble.
To bad that I start to like the Cinnamon 3D desktop. :-) So, which kernels with witch configuration do you
all use?
$ uname -a
Linux chopin 3.17.6-200.fc20.x86_64 #1 SMP Mon Dec 8 15:21:05 UTC 2014
x86_64 x86_64 x86_64 GNU/Linux
Run jack at 64/2 with the ice1712 (delta1010 or 2496) both with
pulseaudio and without. Sporadic xruns when switching workspaces or
with pulseaudio clients entering and exiting quickly but during actual
recording and playback: rock solid.
Running:
- Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz
- 32GB RAM
- 4x 120GB SSD (hardware RAID5)
- 2x 120GB SSD (software RAID0)
- Intel Graphics and Onboard sound (sometimes I give that to
pulseaudio)
8:14 a.m.
On Fri, 23 Jan 2015 20:53:21 -1000, david wrote:
No desktop environment should require hardware
acceleration to work.
That way lies the Land of Windows Bloat!
Amen!
JFTR my stone age ATI graphics, when using the FLOSS driver, can be used
without any performance issues for Google Earth, it's quick as a flash,
but using Cinnamon is impossible. That's ridiculous! Cinnamon and
similar desktops are simply written by people who don't know how to
program correctly. It's not a driver acceleration issue, if it's
possible to use software such as Google Earth without performance
issues, but using a DE as Cinnamon is impossible.
8:40 a.m.
Am 24.01.2015 um 08:14 schrieb Ralf Mardorf:
On Fri, 23 Jan 2015 20:53:21 -1000, david wrote:
Seems like you are right, no problem/Xrun with Kernel: 3.19.0-rc5
x86_64 when I use LXDE instead of Cinnamon.
So it seems that, as soon acceleration in the nouveau driver becomes
improved, Cinnamon isn't suitable for Audio work anymore.
No desktop environment should require hardware
acceleration to work.
That way lies the Land of Windows Bloat!
Amen!
JFTR my stone age ATI graphics, when using the FLOSS driver, can be used
without any performance issues for Google Earth, it's quick as a flash,
but using Cinnamon is impossible. That's ridiculous! Cinnamon and
similar desktops are simply written by people who don't know how to
program correctly. It's not a driver acceleration issue, if it's
possible to use software such as Google Earth without performance
issues, but using a DE as Cinnamon is impossible.
8:53 a.m.
On Sat, 24 Jan 2015 08:39:57 +0100, Hermann Meyer wrote:
Am 24.01.2015 um 08:14 schrieb Ralf Mardorf:
You should run Google Earth on LXDE. Likely the acceleration for your
graphics provided by nouveau is perfect for such advanced 3D software.
Keep in mind that Google Earth allows quick moves from space to street
views, it not only does stress the 3D acceleration, but at the same time
it does stress the Internet connection.
As soon as a desktop environment/window manager needs more resources
for opening or closing a window, than Google Earth does need to move
from space to a street, IMO the DE/WM is broken.
On Fri, 23 Jan 2015 20:53:21 -1000, david wrote:
Seems like you are right, no problem/Xrun with Kernel: 3.19.0-rc5
x86_64 when I use LXDE instead of Cinnamon.
So it seems that, as soon acceleration in the nouveau driver becomes
improved, Cinnamon isn't suitable for Audio work anymore. No desktop environment should require hardware
acceleration to
work. That way lies the Land of Windows Bloat!
Amen!
JFTR my stone age ATI graphics, when using the FLOSS driver, can be
used without any performance issues for Google Earth, it's quick as
a flash, but using Cinnamon is impossible. That's ridiculous!
Cinnamon and similar desktops are simply written by people who
don't know how to program correctly. It's not a driver acceleration
issue, if it's possible to use software such as Google Earth
without performance issues, but using a DE as Cinnamon is
impossible.
9:08 a.m.
Am 24.01.2015 um 08:53 schrieb Ralf Mardorf:
On Sat, 24 Jan 2015 08:39:57 +0100, Hermann Meyer
wrote:
Didn't know what the cause is, but CPU usage of Cinnamon never goes
above 10% , even if I stress a lot with window moving pop-up/down, etc.
Also MEM usage is with more then 700MB (firefox, evolution, geany, etc.
opened) moderate.
With kernel 3.4.xxx-rt there are no problems/Xruns at all.
Broken or not, I'm not as fast as you, to diss a Application/DE or what
ever. Sure there is something wrong in Cinnamon, but, I'm sure that it
could solved, as soon one find out what it is.
Otherwise, Cinnamon is a nice looking Desktop, with nice effects and 3D
impression on the screen.
For now I know, that if I need to switch to never kernels, I've to
switch back to LXDE or a similar lightweight desktop.
regards
hermann
Am 24.01.2015 um 08:14 schrieb Ralf Mardorf:
You should run
Google Earth on LXDE. Likely the acceleration for your
graphics provided by nouveau is perfect for such advanced 3D software.
Keep in mind that Google Earth allows quick moves from space to street
views, it not only does stress the 3D acceleration, but at the same time
it does stress the Internet connection.
As soon as a desktop environment/window manager needs more resources
for opening or closing a window, than Google Earth does need to move
from space to a street, IMO the DE/WM is broken. On Fri, 23 Jan 2015 20:53:21 -1000, david wrote:
Seems like you are right, no problem/Xrun with Kernel: 3.19.0-rc5
x86_64 when I use LXDE instead of Cinnamon.
So it seems that, as soon acceleration in the nouveau driver becomes
improved, Cinnamon isn't suitable for Audio work anymore. No desktop environment should require hardware
acceleration to
work. That way lies the Land of Windows Bloat!
Amen!
JFTR my stone age ATI graphics, when using the FLOSS driver, can be
used without any performance issues for Google Earth, it's quick as
a flash, but using Cinnamon is impossible. That's ridiculous!
Cinnamon and similar desktops are simply written by people who
don't know how to program correctly. It's not a driver acceleration
issue, if it's possible to use software such as Google Earth
without performance issues, but using a DE as Cinnamon is
impossible.
9:28 a.m.
On Sat, 24 Jan 2015 09:08:52 +0100, Hermann Meyer wrote:
Didn't know what the cause is, but CPU usage of
Cinnamon never goes
above 10% , even if I stress a lot with window moving pop-up/down,
etc. Also MEM usage is with more then 700MB (firefox, evolution,
geany, etc. opened) moderate.
It's not important what top does show us, if we experience issues with
GTK3 based, GNOME shell based DEs.
Btw. I even don't trust the new top, not really new, anyway, on Arch I
don't use it, I'm using
[rocketmouse@archlinux ~]$ pacman -Qo /usr/bin/top
/usr/bin/top is owned by procps-ng-classic 3.3.10-2
instead.
https://aur.archlinux.org/packages/procps-ng-classic/
Btw. since you mentioned Evolution, I migrate from Evolution to
Claws, because Evolution became to slow and buggy.
28 Jan
28 Jan
9:04 a.m.
Quoting Hermann Meyer <brummer-(a)web.de>de>:
Am 24.01.2015 um 08:14 schrieb Ralf Mardorf:
I don't think it's Cinnamon or nouveau per se.
When I use Compiz/XFCE and the NVIDIA binary driver I also get xruns
even with fairly high latency settings on a stock 12.04 Ubuntu kernel.
Switching to xfwm4 allows me to go much lower.
To me it seems something in the compositing is the main culprit for
killing low latency.
- Peder
On Fri, 23 Jan 2015 20:53:21 -1000, david wrote:
Seems like you are right, no problem/Xrun with Kernel: 3.19.0-rc5
x86_64 when I use LXDE instead of Cinnamon.
So it seems that, as soon acceleration in the nouveau driver becomes
improved, Cinnamon isn't suitable for Audio work anymore. No desktop environment should require hardware
acceleration to work.
That way lies the Land of Windows Bloat!
Amen!
JFTR my stone age ATI graphics, when using the FLOSS driver, can be used
without any performance issues for Google Earth, it's quick as a flash,
but using Cinnamon is impossible. That's ridiculous! Cinnamon and
similar desktops are simply written by people who don't know how to
program correctly. It's not a driver acceleration issue, if it's
possible to use software such as Google Earth without performance
issues, but using a DE as Cinnamon is impossible.
4 Feb
4 Feb
5:37 a.m.
Am 28.01.2015 um 09:04 schrieb Peder Hedlund:
Quoting Hermann Meyer <brummer-(a)web.de>de>:
Well, but it seems to be kernel related as well, Kernel 3.4.105 didn't
case Xruns, while compositing is on here.
regards
hermann
Am 24.01.2015 um 08:14 schrieb Ralf Mardorf:
I don't think it's Cinnamon or nouveau per se.
When I use Compiz/XFCE and the NVIDIA binary driver I also get xruns
even with fairly high latency settings on a stock 12.04 Ubuntu kernel.
Switching to xfwm4 allows me to go much lower.
To me it seems something in the compositing is the main culprit for
killing low latency.
- Peder
On Fri, 23 Jan 2015 20:53:21 -1000, david wrote:
Seems like you are right, no problem/Xrun with Kernel: 3.19.0-rc5
x86_64 when I use LXDE instead of Cinnamon.
So it seems that, as soon acceleration in the nouveau driver becomes
improved, Cinnamon isn't suitable for Audio work anymore. No desktop environment should require hardware
acceleration to work.
That way lies the Land of Windows Bloat!
Amen!
JFTR my stone age ATI graphics, when using the FLOSS driver, can be
used
without any performance issues for Google Earth, it's quick as a flash,
but using Cinnamon is impossible. That's ridiculous! Cinnamon and
similar desktops are simply written by people who don't know how to
program correctly. It's not a driver acceleration issue, if it's
possible to use software such as Google Earth without performance
issues, but using a DE as Cinnamon is impossible.
3378
days inactive
3390
days old
linux-audio-user@lists.linuxaudio.org
32 comments
12 participants
participants (12)
-
Atte -
david -
David Santamauro -
Harry van Haaren -
Hermann Meyer -
Jeremy Jongepier -
Len Ovens -
Peder Hedlund -
Raffaele Morelli -
Ralf Mardorf -
rob -
Will Godfrey