Audiolinux Server configurations, Software, Hardware, and Listening Impressions

[drjimwillie]

 Thank you for the quick response I will try it.   I will try rebooting and reloading first. 

 But, if it does not work I do not know how to find directories to find the .config file in Linux. 

 Will you please look at my other screenshots to see if my settings look OK. 

 What is one time network time service? 

 What is one time I RQ thread? 

 I have never seen these two lines before, are they new? 

 The first screen shows my CPU speed has 3.9 GHz max.   Recommended 3.5 GHz max.  Is that what CPU speed 3500.009 is?

 Is that correct for boot mode standard? 

[drjimwillie]

 I loaded the new kernel and the ACPI driver. 

 At first I thought it was going to take me a minute to learn how to hear this new sound.   There’s more energy to it, like pushing out a strong breath.

 I’m not playing the music that impresses me technically, I’m playing my old favorites to see if they give me the emotional response like they used to when I first heard them  on a record.

[luisma]

On 6/14/2019 at 8:34 AM, hifi25nl said:

3 is for CPU0 and CPU1, if all other cores are isolated (CPU2 CPU3 etc.)

 

If you have all cores less the first (CPU0) isolated, the right value should be 1 

Hello Everyone, I'm trying to optimize my AL config following this thread, I have followed Piero's comments and others but still need some help.

 

ENABLING CORES FOR AUDIO (italics are actual text from AL)

How many cores do you want to use exclusively for audio?
1) All but the first
2) Half
3) Only one (random, excluding first)
4) Disable

I select 1 as I would like to use 3 cores for audio and 1 for system

But it shows this

ISOLATED CPU CORES FOR AUDIO: 1

Which confuses me as I understand only ONE core for audio, which statement is correct in this case?

And even more confusing is the cpumask output which is not a number but an F

[audiolinux@alsvr ~]$ cat /sys/bus/workqueue/devices/writeback/cpumask
f

 

ENABLING REALTIME EXPERT CONFIG

This specific option is on the default, what is your experience with it? leave it on or off?
echo 0 >/sys/devices/system/machinecheck/machinecheck3/check_interval

 

Thanks

[hifi25nl]

Did you reboot? Configuration is effective only after a reboot.

[luisma]

Yes I did, this is the menu output

 

cpumask remains at f

 

Edit: ramroot is disabled as I don't have enough memory

 

[hifi25nl]

How many physical cores do you have? If you have only 2 this is correct.

Isolated cores will disable hyper-threading so if you have core0 and core1 only, if you select all but the first, you should have only the second core (1) isolated.

 

If you have more than 2 physical cores, for example 4, please check the file /boot/loader/entries/audiolinux.conf

You should have isolcpus=1-3 in the kernel line

[luisma]

24 minutes ago, hifi25nl said:

How many physical cores do you have? If you have only 2 this is correct.

Isolated cores will disable hyper-threading so if you have core0 and core1 only, if you select all but the first, you should have only the second core (1) isolated.

 

If you have more than 2 physical cores, for example 4, please check the file /boot/loader/entries/audiolinux.conf

You should have isolcpus=1-3 in the kernel line

Thanks Piero, of course HT is disabled, only 2 cores so 1 used for audio

 

It is correct, thanks for the assistance

 

Oh this option

echo 0 >/sys/devices/system/machinecheck/machinecheck3/check_interval

Is it recommended?

 

 

[hifi25nl]

I would try all...:

 

echo 1000 > /proc/sys/vm/stat_interval
echo 0 > /sys/bus/workqueue/devices/writeback/numa
echo 1 > /sys/bus/workqueue/devices/writeback/cpumask
echo -1 > /proc/sys/kernel/sched_rt_runtime_us
echo 0 >/sys/devices/system/machinecheck/machinecheck3/check_interval
echo 0 > /proc/sys/kernel/watchdog
echo 0 > /proc/sys/kernel/nmi_watchdog

echo 1500 > /proc/sys/vm/dirty_writeback_centisecs
 

The line echo 1 > /sys/bus/workqueue/devices/writeback/cpumask should be used only in your case (general purpose cpu equivalent to CPU0)

[luisma]

2 minutes ago, hifi25nl said:

I would try all...:

 

echo 1000 > /proc/sys/vm/stat_interval
echo 0 > /sys/bus/workqueue/devices/writeback/numa
echo 1 > /sys/bus/workqueue/devices/writeback/cpumask
echo -1 > /proc/sys/kernel/sched_rt_runtime_us
echo 0 >/sys/devices/system/machinecheck/machinecheck3/check_interval
echo 0 > /proc/sys/kernel/watchdog
echo 0 > /proc/sys/kernel/nmi_watchdog

echo 1500 > /proc/sys/vm/dirty_writeback_centisecs
 

The line echo 1 > /sys/bus/workqueue/devices/writeback/cpumask should be used only in your case (general purpose cpu equivalent to CPU0)

Perfect, thank you so much for clarifying

 

 

[flkin]

23 hours ago, hifi25nl said:

I would try all...:

 

echo 1000 > /proc/sys/vm/stat_interval
echo 0 > /sys/bus/workqueue/devices/writeback/numa
echo 1 > /sys/bus/workqueue/devices/writeback/cpumask
echo -1 > /proc/sys/kernel/sched_rt_runtime_us
echo 0 >/sys/devices/system/machinecheck/machinecheck3/check_interval
echo 0 > /proc/sys/kernel/watchdog
echo 0 > /proc/sys/kernel/nmi_watchdog

echo 1500 > /proc/sys/vm/dirty_writeback_centisecs
 

The line echo 1 > /sys/bus/workqueue/devices/writeback/cpumask should be used only in your case (general purpose cpu equivalent to CPU0)

 

If I Enabled Only One core for music in my 8 core CPU (random but say core 3 only), would the offset for

 

echo x > /sys/bus/workqueue/devices/writeback/cpumask

 

be x = 247?

 

ie.  In the web link you posted above to calculate the offset, we should be selecting the NOT FOR MUSIC cores? So for Enabled Only One core 3, we select 1,2,4,5,6,7,8 for the calculation?

 

Since the cores are random each time a power cycle occurs, does that mean we have to edit this offset and enter the proper number again should the random core switch?

[hifi25nl]

If you want only one core isolated , after choosing random, you can edit isolcpus in /boot/loader/entries/audiolinux.conf with your preferred core.

For example: isolcpus=7 

This will stick after a reboot (random does not mean that will change each time)

 

When using the calculator you must be sure that Hyper-Threading is disabled.

For example, if you choose only core 7 (core 8 in the calculator) the cpumask number should be 128

[flkin]

Got it, thanks. Entering the incorrect mask number affects the sound quite a bit!

[rickca]

8 hours ago, hifi25nl said:

When using the calculator you must be sure that Hyper-Threading is disabled.

What calculator are you guys talking about?

[elan120]

2 minutes ago, rickca said:

What calculator are you guys talking about?

This calculator:  http://store.tomandmiu.com/cal-cam.php

[Gavin1977]

Over the last few days Piero has been helping me resolve an issue of no sound in Audiolinux lxqt.  I now have audio working through all desktop applications (I followed this guide).  I believe the issue just affected applications used within lxqt destop, like chromium or local players such as HQ Player.  If I were using Roon, then reading the comments on this forum it would have auto-configured and have been a non-issue.  Why this had to be done manually I do not know, as confusingly the configuration screen for the sound card in Alsa Mixer always showed up as being correct... however it has resolved the issue.  I hope this might help save others from the frustration!

 

Subjectively I would rank performance something like the following (Config: DNKE NUC, 16GB RAM, OPTANE H10 and Naim DAC V1):

 

BEST

Euphony

Audiolinux (RAM Boot / Extreme)

.

.

.

.

Roon Rock / Audiolinux (Boot from USB)

.

MacBook Pro

WORST

 

So what I have gathered from this is that operating in RAM really makes quite a difference and backs up what others have said.  I believe the next steps for improving my Audiolinux setup would be to run it headless (it was nice to work with lxqt to start with), and probably as a squeezelite or Roon bridge.

 

I should also mention that I'm, just using the stock SMPS brick that came with the NUC, I can still hear significant improvements with Euphony and Audiolinux Ramboot.  I would say that I am satisfied with the sound of Euphony or Audiolinux RAMBoot compared to the commercial streamers/servers I have heard even with this SMPS.  My Paul Hynes SR5 will be with me in September (touch wood), so I will be excited to hear what it brings to the table.

 

I feel as though the slightly richer sound of Euphony (the remaining differentiating factor from Audiolinux RAMboot by what I can ascertain) can probably be matched by further tweaks I have heard about on this forum, such as:

 

 - isolate cores / CPU frequency adjustments

 - removal of surplus drives... is it best best to run from USB Flash Drive to memory only (i.e. no Optane and no 2.5" disk)?

 

I have no experience with the above - what configs should I implement next?

[rickca]

@hifi25nl I see that the Ryzen 9 (3950X 16 cores and 3900X 12 cores) are actually two chiplets, so the 3950X is 8+8 and the 3900X is 6+6.  Do these CPUs require NUMA support in AL?

[hifi25nl]

I think they should work both ways, but it is early to give a complete answer.

What would be very interesting is to use half of the cores for audio, half for general purposes. With NUMA you can divide memory and cores. A sort of "Audiolinux twins"...

 

 

[Nenon]

On 7/4/2019 at 9:33 AM, Gavin1977 said:

is it best best to run from USB Flash Drive to memory only (i.e. no Optane and no 2.5" disk)?

 

IMO - Yes. And remove the USB Flash Drive after AudioLinux loads in RAM. 

But in my system, I don't hear any difference if loading to memory from Optane instead of USB Flash Drive. The data transfer to memory, essentially the booting, is faster this way. 

[Gavin1977]

12 hours ago, Nenon said:

 

IMO - Yes. And remove the USB Flash Drive after AudioLinux loads in RAM. 

But in my system, I don't hear any difference if loading to memory from Optane instead of USB Flash Drive. The data transfer to memory, essentially the booting, is faster this way. 

Thanks - I'll make that my next port of call.

[RDOK]

Hello all - I am new on this forum.

Enthusiasted by the long thread elsewhere on this forum I installed a humble nuc7p (on an LPS) as my Roon bridge.

Running AL on it and from there straight by USB (good home made cable) into my Devialet amp.

Until recently the sound quality was almost undistinguishable from the wifi or ethernet connection from the Roon core to the Dev.

The latest addition on AL, moving from p-state to ACPI, made a small but very significant step upward in the sq.

The sound now has obtained a more 3D chracter; more depth not (only) in soundstage but around the instruments and voices.

Once heard I definitely want to keep this and not go back.

Obviously this is just my personal experience, with my ears in my system; not more, not less.

But the change in sound perception this last step did for me is so rewarding it motivated me to post this as feedback on AL.

 

Thanks Piero, great work!

[AnotherSpin]

On 6/21/2019 at 8:36 PM, elan120 said:

 

To change from intel pstate to acpi, try the following:

 

 

• Open menu and change ramroot from enable to disable.
• Re-boot, make sure it is boot to Audiolinux in boot menu (top choice) during boot process.
• Go to menu selection “1” - “Configuration”.
• Select “0” (Change scaling driver to acpi or intel pstate) and follow with “Yes” when new screen come up to change from intel pstate to acpi.  Note, only do this once to make scaling driver change to acpi; if done twice, it will change from acpi back to intel pstate.
• Re-boot, check system menu, now scaling driver should be acpi.
• Re-enable ramroot.

 

For changing to acapi in headless AL I do the following:

 

- Configuration > 19. Disable ramroot

- Save

- Reboot

- Configuration > 21. Set CPU governor (...)

- 0. Change scaling driver to acpi (...)

- reboot

 

ACPI status shown on the top of first page, ramroot would be enabled back as usual.

 

[ferenc]

I purchased and installed AL few days ago in my Asus P10-WS, 4 core 25W TDP Xeon audio pC. Sounds really nice with JPlay Femto USB card and Holo Spring 2 DAC and Singxer SU-1. I got an Audio Note/EVGA Nu Audio PCIe sound card as well. Officially it is Windows only. However it uses an Xmos USB chip internally as a "virtual" USB device. I am using the audio PC as Roon Bridge, Roon running in my iMac Pro. Roon sees the Nu Audio board as 2 different cards. One way it sounds distorted and the other way it is few decibel quite than the level actually is, and this way DSD converted automatically to PCM which is strange as I do not use Roon to convert DSD to PCM. So my question is there any way to get the Nu Audio board working properly in AL? Any suggestion what to check, change? Thanks in advance. 

[lmitche]

On 7/11/2019 at 2:18 AM, ferenc said:

I purchased and installed AL few days ago in my Asus P10-WS, 4 core 25W TDP Xeon audio pC. Sounds really nice with JPlay Femto USB card and Holo Spring 2 DAC and Singxer SU-1. I got an Audio Note/EVGA Nu Audio PCIe sound card as well. Officially it is Windows only. However it uses an Xmos USB chip internally as a "virtual" USB device. I am using the audio PC as Roon Bridge, Roon running in my iMac Pro. Roon sees the Nu Audio board as 2 different cards. One way it sounds distorted and the other way it is few decibel quite than the level actually is, and this way DSD converted automatically to PCM which is strange as I do not use Roon to convert DSD to PCM. So my question is there any way to get the Nu Audio board working properly in AL? Any suggestion what to check, change? Thanks in advance. 

Hi Ferenc,

 

It's been a week since your post and no one has responded. I for one have never heard of the Nu Audio board. Did you ever get this resolved?

 

Does the virtual USB port show up when you type lsusb into the command line?

 

Larry

[ferenc]

6 hours ago, lmitche said:

Hi Ferenc,

 

It's been a week since your post and no one has responded. I for one have never heard of the Nu Audio board. Did you ever get this resolved?

 

Does the virtual USB port show up when you type lsusb into the command line?

 

Larry

 

Yes, with some help from Piero, I got it work in PCM. The two cards Al sees is the Playback and Capture part of the card, obviously I do not need its capture part at the moment. So it words well with Roon as a Bridge an with HQP as NAA endpoint. 

 

The DSD playback does not work, not with DoP, not natively, only if I convert the DSD files to PCM in Roon or in HQP.

 

So for the proper DSD playback there is a need a Kernel upgrade it seems. Using Win 2019 server or Win 10 the Nu Audio card plays my 2X HD DSD256 files really well. 

 

The Nu Audio Note / EVGA Nu Audio card sounds really nice, I just did not find any fault in its sounds, very balanced, smooth and the music nicely flows. It is a steal for the price. The PC needs a good power supply and somehow the card a bit sensitive on the interconnect cables. But this is it. If one wants a simple, straightforward very high quality audio PC, with no fault, Audiolinux works perfectly well in PCM with this card and you will not need any magic with USB cards, cables, any USB gadgets.

 

My audio PC is a completely passive PC, with ASUS P10-WS workstation motherboard, with 4 core low power (25W TDP) Xeon, 16 (4x4) GB ECC RAM an external 20V, 10A linear power supply with HDplex 400W DC-ATX converter. Sounds perfectly fine for me. Roon and HQP is running on an 18 core iMac Pro. All switches, router, set top boxes are powered with LPS. Cabling is mostly Wireworld Platinum, including power cables.

[luisma]

On 7/4/2019 at 10:33 AM, Gavin1977 said:

BEST

Euphony

Audiolinux (RAM Boot / Extreme)

.

.

.

.

Roon Rock / Audiolinux (Boot from USB)

.

MacBook Pro

WORST

Pardon me asking, when you did this comparison from BEST to WORST did you followed the same exact testing for all OS's and conditions? meaning consistently using the same player (from your 3rd line) in this case ROON? I'm asking because to me these is a difference between ROON playing and STYLUS playing in Euphony (I prefer Stylus) and would like to know your opinion. Thanks