Xruns with Audiophile 2496 M-audio
Hello,
I read a lot a great thing about the audiophile 24/96 on the net, and i decided to buy it to get better latency than my old sblive! I just received the audiophile 24/96 and unfortunatly, i have a lot of XRuns with when i run jackd.
I already tried the following :
- i changed irq (several changes : irq5, then irq 10)
- Disable nearly all the inboard harware (serial port, parallel port, usb)
- i changed the Pci port
- i tried to run jackd as root
- i even tried to raise buffer at big values : 512, 1024, and even there i have some (1 or 2) xruns
- i'm now trying with kernel 2.6.6 to see if things are better
Hardware :
MSI K7N2Delta MB
Athlon 2400
Nvidia Geforce 4mmx
Audiophile 24/96 M-Audio
Sofware :
Debian/Testing
Kernel 2.6.5
Alsa 1.0.4rc2
Realtime-Lsm 0.1.0
Jackd 0.98
Do you have ideas, before i return the card to the shop ?
Regards,
Rémi Bernhard.
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_id�66&op=click
[Remi Bernhard]
Quoted 4 lines
I read a lot a great thing about the audiophile 24/96 on the net, and i decided to buy it to get better latency than my old sblive! I just received the audiophile 24/96 and unfortunatly, i have a lot of XRuns with when i run jackd.
my old system had various 2.4.x-ll kernels, asus board and a k6-III 450, then migrated to kernel 2.4.19-ll, a cheap-as-hell board (forgot the brand name) and an athlon xp 2k, ALSA versions ranging from pre-0.9 CVS to 1.0.3.
the PCI audiophile 24/96 ran and runs smooth as a train in all configurations at 64 frames / 44.1 kHz.
obvious question, is your kernel ll-patched? if not, go here:
http://www.zipworld.com.au/~akpm/linux/schedlat.html
(i gather that 2.4 is still the better choice for low-latency operations.)
oh, btw, the ISA 'Blaster AWE64 i used to run in the old board did 64/44.1 just as stable as the audiophile.
cheers,
tim
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
Hi,
Quoted 9 lines · On Sat, 29 May 2004 12:52:15 +0200 (CEST) Tim Goetze <[email protected]> wrote:
[Remi Bernhard]
Quoted 4 lines
I read a lot a great thing about the audiophile 24/96 on the net, and i decided to buy it to get better latency than my old sblive! I just received the audiophile 24/96 and unfortunatly, i have a lot of XRuns with when i run jackd.
obvious question, is your kernel ll-patched? if not, go here:
As i said ,
<rb>
Sofware :
Debian/Testing
Kernel 2.6.5
</rb>
So i have 2.6.5 kernel, and not 2.4.x. Do you think i should install 2.4.25 instead ?
Quoted 4 lines
(i gather that 2.4 is still the better choice for low-latency operations.)
I read the alsa-* archives, and i saw a thread with two people who had good latency results with 2.6.x.
Could someone tell me if man can achieve a buffer size <=128 with audiophile 2496 and kernel 2.6.x ?
Quoted 2 lines
oh, btw, the ISA 'Blaster AWE64 i used to run in the old board did 64/44.1 just as stable as the audiophile.
Great !
Regards,
Remi.
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
Quoted 4 lines · Remi Bernhard
So i have 2.6.5 kernel, and not 2.4.x. Do you think i should install 2.4.25 instead ?
Quoted 1 line
Quoted 9 lines
Quoted 3 lines
(i gather that 2.4 is still the better choice for low-latency operations.)
I read the alsa-* archives, and i saw a thread with two people who had good latency results with 2.6.x.
Could someone tell me if man can achieve a buffer size <=128 with audiophile 2496 and kernel 2.6.x ?
Yes. The 2.4 kernels with low latency patch are better than 2.6 at present (even the mm series). The buffer sizes you can achieve are dependent on so many different things (disks, FSB, memory, configuration...) that there is no way to tell. You just have to try it and find out. As far as minimizing buffer size is concerned - if you are doing live audio or you want to monitor recording with inboard effects turned on you need to have a small buffer size. With the envy24 chipset cards you can use hardware monitoring so you can set the buffer size to 2048 and have zero latency monitoring. Don't confuse monitoring latency with system latency.
Jan
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
Hi,
Quoted 6 lines
As far as minimizing buffer size is concerned - if
you are doing live audio or you want to monitor recording with inboard effects turned on you need to have a small buffer size. With the envy24 chipset cards you can use hardware monitoring so you can set the buffer size to 2048 and have zero latency monitoring. Don't confuse monitoring latency with system latency.
To be more sharp, i need to reduce latency because i use softsynth (zynaddsubfx, fluidsynth, etc.) that needs a low latency when playing from midi keyboard, and if not, there is a "lag" time, that man can hear, between the time when you hit the key, and the time it is played.
I don't know about wich latency i spoke, i think it is system latency. So I think that buffer size is **the** parameter to change.
Is that true ?
Regards,
Remi.
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
[Remi Bernhard]
Quoted 9 lines
To be more sharp, i need to reduce latency because i use softsynth (zynaddsubfx, fluidsynth, etc.) that needs a low latency when playing from midi keyboard, and if not, there is a "lag" time, that man can hear, between the time when you hit the key, and the time it is played.
I don't know about wich latency i spoke, i think it is system latency. So I think that buffer size is **the** parameter to change.
Is that true ?
yes.
tim
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
Quoted 6 lines · Tim Goetze
[Remi Bernhard]
Quoted 5 lines
To be more sharp, i need to reduce latency because i use softsynth (zynaddsubfx, fluidsynth, etc.) that needs a low latency when playing from midi keyboard, and if not, there is a "lag" time, that man can hear, between the time when you hit the key, and the time it is played.
Quoted 7 lines
Quoted 4 lines
I don't know about wich latency i spoke, i think it is system latency. So I think that buffer size is **the** parameter to change.
Is that true ?
yes.
Yes, that is the parameter to change but no, that isn't system
latency. System latency is response to interrupt time not buffer fill time.
Jan
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
[Jan Depner]
Quoted 10 lines · Tim Goetze
Quoted 9 lines
[Remi Bernhard]
Quoted 8 lines
To be more sharp, i need to reduce latency because i use softsynth (zynaddsubfx, fluidsynth, etc.) that needs a low latency when playing from midi keyboard, and if not, there is a "lag" time, that man can hear, between the time when you hit the key, and the time it is played.
I don't know about wich latency i spoke, i think it is system latency. So I think that buffer size is **the** parameter to change.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
Is that true ?
yes.
Yes, that is the parameter to change but no, that isn't system
latency. System latency is response to interrupt time not buffer fill time.
depends on what 'system' we're talking about, which isn't really clear in the first place (nor is it too important, but here goes anyway ... :)
you can see the whole setup as the system, then latency is the time from keypress to voltage change at the DAC out. or you can just look at the kernel side as as you do. or you can look at the time from MIDI interrupt to the audio DAC converting the first affected audio sample.
all examples of valid 'systems' to look at in this context, depending on whether you assume the musician's, the kernel- or the audio application-programmer's view.
sorry for nitpicking.
tim
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
Quoted 4 lines · Tim Goetze
depends on what 'system' we're talking about, which isn't really clear in the first place (nor is it too important, but here goes anyway ... :)
Quoted 9 lines
you can see the whole setup as the system, then latency is the time from keypress to voltage change at the DAC out. or you can just look at the kernel side as as you do. or you can look at the time from MIDI interrupt to the audio DAC converting the first affected audio sample.
all examples of valid 'systems' to look at in this context, depending on whether you assume the musician's, the kernel- or the audio application-programmer's view.
Agreed. But the problem that keeps popping up on the lists is that
people who are not doing live sound, have cards that do hardware monitoring, and don't need to use a tiny buffer size waste their time trying to get the minimum buffer size because they think they need to get 2-3ms latency. So, they get xruns out the wazoo and wonder how to fix it. In this case he *needs* a small buffer size though since he's doing, essentially, a live application.
Jan
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
[Jan Depner]
Quoted 7 lines
Agreed. But the problem that keeps popping up on the lists is that
people who are not doing live sound, have cards that do hardware monitoring, and don't need to use a tiny buffer size waste their time trying to get the minimum buffer size because they think they need to get 2-3ms latency. So, they get xruns out the wazoo and wonder how to fix it. In this case he *needs* a small buffer size though since he's doing, essentially, a live application.
agreed.
there must be a wiki or some other sort of documentation regarding all the latency issues out there. at least it would be nice to hand out an URL that answers the most common questions (they do indeed seem to come up time and again even on the -dev lists).
tim
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
[Remi Bernhard]
Quoted 2 lines
So i have 2.6.5 kernel, and not 2.4.x. Do you think i should install 2.4.25 instead ?
i do think so because i never saw any hard facts about latency posted for the 2.6 series, while 2.4.x-ll is well documented and proven to work quite well here and elsewhere.
it _could_ also be that your graphics subsystem is hogging the PCI bus for long enough to cause xruns, but that is mostly guessing since i never laid my hands on a linux box with nvidia hardware. i take it you're not running graphics-intensive applications while testing the audio subsystem anyway.
as Jan has pointed out, make sure your IDE subsystem is using DMA on all devices, too. (man hdparm)
tim
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
Quoted 7 lines · On Sat, 29 May 2004 13:44:40 +0200 (CEST) Tim Goetze <[email protected]> wrote:
[Remi Bernhard]
Quoted 2 lines
So i have 2.6.5 kernel, and not 2.4.x. Do you think i should install 2.4.25 instead ?
i do think so because i never saw any hard facts about latency posted for the 2.6 series, while 2.4.x-ll is well documented and proven to work quite well here and elsewhere.
I tried with kernel 2.4.25 + lowlatency patch.
I have the same xruns :-/
Any other idea ?
Regards,
Rémi.
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_id�66&op=click
[Remi Bernhard]
Quoted 4 lines
I tried with kernel 2.4.25 + lowlatency patch.
I have the same xruns :-/Any other idea ?
checklist:
* running jackd as root? with -R option?
* tried running jackd without any clients?
* IDE-DMA enabled?
* ll-sysctl interface chosen at kernel build time? turned it on?
* X running? tried without?
* tried 'latencytest' from the ALSA tree?
tim
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
Hi Tim and all,
Quoted 9 lines · On Sat, 29 May 2004 16:34:49 +0200 (CEST) Tim Goetze <[email protected]> wrote:
[Remi Bernhard]
Quoted 4 lines
I tried with kernel 2.4.25 + lowlatency patch.
I have the same xruns :-/Any other idea ?
checklist:
* running jackd as root? with -R option?
Tried -> failed.
Quoted 1 line
* tried running jackd without any clients?
It worked, but as i put some clients, i get xruns again.
Quoted 1 line
* IDE-DMA enabled?
Yes.
Quoted 1 line
* ll-sysctl interface chosen at kernel build time? turned it on?
Yes.
Quoted 1 line
* X running? tried without?
I didn't tried that.
Quoted 1 line
* tried 'latencytest' from the ALSA tree?
I didn't tried that.
I thank you a lot guys for your help and ideas, but now i've done enought twiking / hacking / patching kernel, switching the audiophile from pci port to pci port, changing irq, recompiling kernel 2.6.5, 2.4.25, 2.6.6, patching again, fails again, etc. etc.
Now i have reinstalled my old sblive! with 512 buffer size, and i'm happy with it (for the moment). I think i'll return the audiophile 24/96 next week.
Thanx again for all your help.
Regards,
Rémi Bernhard.
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_id�66&op=click
Quoted 10 lines · Remi Bernhard
Hi Tim and all,
Quoted 4 lines · On Sat, 29 May 2004 16:34:49 +0200 (CEST) Tim Goetze <[email protected]> wrote:
[Remi Bernhard]
Quoted 3 lines
I tried with kernel 2.4.25 + lowlatency patch.
I have the same xruns :-/
Quoted 7 lines
Quoted 5 lines
Quoted 1 line
Any other idea ?
checklist:
* running jackd as root? with -R option?
Tried -> failed.
If this failed, you're not running real-time. What error message did
you get?
Jan
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
Quoted 15 lines · On Sun, 30 May 2004 08:35:21 -0500 Jan Depner <[email protected]> wrote:
Quoted 14 lines · Remi Bernhard
Hi Tim and all,
Quoted 8 lines · On Sat, 29 May 2004 16:34:49 +0200 (CEST) Tim Goetze <[email protected]> wrote:
[Remi Bernhard]
Quoted 4 lines
I tried with kernel 2.4.25 + lowlatency patch.
I have the same xruns :-/Any other idea ?
checklist:
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
* running jackd as root? with -R option?
Tried -> failed.
If this failed, you're not running real-time. What error message did
you get?
Oops, sorry, my mistake. **This** worked.
Regards,
Rémi Bernhard.
-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id149&alloc_id�66&op=click