Subjective Impressions of Sound Quality - Audirvana Studio on Linux

I listened to it on a MacBook Pro and the Arch Linux minimal install. I didn’t turn it up loud enough to hear the extraneous noises. :slightly_smiling_face:

Of course as always this is purely a subjective impression in my own system, however reliable or unreliable that may be.

I felt the instruments had more body and there was more sense of space hearing it with the Arch Linux minimal install. There was also a better sense of reverberation and decay, so that the timing of the interplay between instruments was more evident.

Both the MacBook Pro and the PC have CPUs that are very capable, so I think any difference must come from how few services are running on the Arch Linux minimal install.

In your case do the CPUs perhaps have differing capabilities?

Many thanks Jud for your feedback! Appreciated!
What you hear is quite interesting especially about the feeling of body of instrument which is essential in this piece, especially in that andante. There is space between the two mandolins, and a bit of reverberation that gives you the idea of the recording place (Most probably a big room, with a window on the left side, wide open due to a hot summer somewhere in Italy…).

In my case, there is indeed a big difference between the two CPUs: an Intel Core i3 for the Mac Mini, with MAC OS Ventura, and a RealTek 1296 (ARM processor with 4 cores).

I would be interested to try an installation of your Arch Linux on the Mac Mini or on a MacBook Pro (I have one, a 2012 version I guess).
Is it simple to install that Linux thing in a Mac with BootCamp?

Anyway, what you proved in a way is that:

  • A system with a small memory and CPU footprint is better from an audio point of view,
  • using fiber ethernet is superior to copper. This is also something I should try…
1 Like

I would suggest if you look at fiber, that it be at least 10G (SFP+). You don’t need the bandwidth, but the 10G specification requires not only low jitter, but that if the interface receives a higher jitter signal, it will reduce the jitter to the required level. And it is not expensive. Here’s the switch I use:

I haven’t attempted it, and have not looked to see if others have written about it, so I would not want to speculate. I can tell you it is not the most simple installation I have done even on a PC. :slightly_smiling_face:

Thanks for the link. I was thinking about this fiber solution, but I do not have any fiber input to my Mac mini or streamer, so I would need to install the fiber link between the Mac and the streamer, and this will require two switches with SFP+ slots.
If I’m not totally mistaken, of course.
In my setup, I have:

  • On one side, a Mac Mini that hosts AS.
  • A Syno diskserver that hosts the audio files
  • On the other side, a streamer connected to a DAC via an AES/EBU link. And then the rest of the HIFI system.

If I use a fiber link, it will be between the Mac and the streamer, but at the end on both sides, this will require a switch with SFP+. Why not but would I gain something if I introduce another intermediate equipment (the switch) and additional Ethernet copper cables?

Of course, the MAC needs also to be connected to the LAN and to Internet… And that will require the switch with SFP+ slot to be connected to them as well.

Not really simple. Once again, if I’m not mistaken.

Yes, in your case it seems you would need multiple boxes, and more complexity is something I don’t favor. Here I have optical Ethernet in the walls and a UPnP endpoint that accepts optical Ethernet input. The only reason for the switch is to supply other equipment (a WiFi router, an Apple TV) with copper Ethernet connections. So there are no extra boxes beyond what I would need anyway for a copper connection.

That’s pretty clear Jud. Now I understand why the use of optical Ethernet is a perfect solution for you.
I will envisage that solution when I am able to deploy a full optical solution at home.
First point will be to have an ethernet router (Box from my ISP) equipped with an output port with an optical fiber: not the case today. Then, I will be able to deploy optical connection to the different rooms. Not for today I presume, and not even for tomorrow.

1 Like

We are still waiting for the ISP to provide optical Ethernet here as well. :slightly_smiling_face: I just put another one of those switches where the service comes into the house, so it supplies the optical fiber throughout the entire home.

@patifr … The only thing proven here, is there is a preference that is intrinsically tied to the user system and the biased interpretations of the listener… What is being shown here, is a problem with semantically and connotative, contextual interpretation being a weak methodology for discernment of sound-quality, in the context of forum posts…

:notes: :eye: :headphones: :eye: :notes:

Yes, maybe… But, what is clear to me (and to several people) is the fact that differences are clearly heard when you use different setup while keeping the same Digital to Analog chain. When we identify differences, we are not judging the quality of what we hear, we simply say that we hear something different.
So, to those who regularly repeat “Bits are bits, cables are all equal”, I would say “Why not trying to open you ears and the rest?”.
But anyway, even if we are manipulated by powerful Gurus who betwitched us, and if we are 100% wrong, what’s the point? Is this really an “Highway to Hell”? Great!


Yes, I often remark that these impressions are purely my subjective preference in my own system (though there are objective measurements by others backing up several of the choices I’ve made). We are not splitting the atom here, simply enjoying having a conversation about music and its reproduction. :+1:


I am thoroughly impressed with how good AS on Linux sounds.

As another comparison experiment, I compared running AS (as server/player/everything!) on the same hardware to a lightweight Roon bridge endpoint and HQ player NAA endpoint.
In my subjective impression, AS wins, followed very closely by the NAA endpoint and then at some distance followed by Roon bridge.
Hats off to AS for doing this standalone in system, vs heavylifting serverwork for Roon and HQplayer being done in a relatively powerful iMac elsewhere in the network.

The used hardware is a Intel N97 based SBC, running from a linear power supply with optical ethernet to isolate the network nasties nicely away.
Installed on it is Arch Linux 6.9.2

Full disclosure: I have decided to build this into a product and plan to introduce it on Pacific Audiofest Seattle in September…

Cheers, Hans.


Good luck to you, Hans! :slightly_smiling_face:

1 Like

Hi @hlooman,

Feel free to reach out if you need anything on my end, like logos, for example :wink:


Could you elaborate a bit on what you did? E.g. AS Linux straight into your system versus AS Linux into Roon Bridge/HQPlayer NAS on another SBC feeding your system?


Currently Audirvana Windows with UPnP and Diretta (with last improovment) work better than Audirvana Linux and UPnP. I currently have no enough hardware to do Audirvana Linux UPnP and Diretta. But I tried with WSL2 and it sound better but so painfull (Firewall, network instability) So waiting with hope Audirvana Linux with Diretta…

I did 3 tests, using the same hardware.
The hardware is always the SBC connecting to my DAC via USB. The music always comes in via a fiber media converter into the ethernet port of the SBC. The SBC and fiber media converter are powered from a linear power supply.
I just changed what’s running on the SBC in always the same Linux, with no other processes running than the setup requires to play, i.e. when Roon Bridge is running AS is completely turned off.

Test 1: AS Linux on SBC straight to the DAC, no other player server needed.
Test 2: Roon Bridge Linux on SBC to the DAC, Roon core running on iMac connecting to SBC via optical ethernet.
Test 3: HQPlayer NAA endpoint on Linux on SBC to the DAC. Roon core running on iMac playing via HQPlayer on same iMac to SBC via optical ethernet.

In test 1 the SBC is the player and server and endpoint all in one.
In tests 2 and 3 the SBC is only used as an endpoint.

It is interesting to me how I still like the sound quality in test 1 best, despite it running the heaviest process (I think) of all 3. Probably the approach of running music from a large buffer straight into the DAC beats using endpoints with smaller buffers and some kind of needed extra transport protocol.
Usually the idea of using small endpoints is being able to run those in light weight mode from linear power supplies, where the player is located in a much noisier computer environment. In test 1 the SBC IS a low noise computer environment able to run AS as player.

In all cases I used no upsampling, playing just bitperfect. The difference in sound quality between tests 2 and 3 seem to come down to the transport protocol being different. Roon Raat vs HQPlayer Naa. I like Naa better. Or perhaps the Naa endpoint handles audio differently than the Roon bridge.
To me, AS wins, followed by Naa and Roon last.

Just my subjective 2 pennies here.

Cheers, Hans.


I recall comparisons with Taiko Extreme via USB output vs Taiko Extreme to a Sonore opticalrendu as endpoint also via USB output. The Taiko USB direct output was much superior also despite running the heaviest process. It seems that computer with higher powered CPUs sound better than these with lower powered ones. So these low powered endpoint solutions do not make much sense to me.

I think you’ll find opinions vary regarding which sounds best to different people in different systems. :slight_smile: Lots of people prefer direct connection, and that’s great.

Please also note Hans’ SBC for the direct connection test uses a 4-core lower-powered (TDP of 12 watts, though up to 3.6GHz) CPU.

1 Like

I have a Sonore optical rendu laying around here catching dust because it never sounded close to the quality I am looking for.
The Taiko is very powerful but very expensive because of it being a full blown PC that has been optimized in all aspects for audio performance. The Taiko SHOULD win being an endpoint compared to a Sonore that costs 30 x less…

My little SBC is actually also a full blown PC, it can easily run Windows, like the Taiko. The nice thing about it is that its power consumption is low enough to allow passive cooling and linear power supplies feeding it.
It’s running a steady 0.82 A at 17VDC so around 14W in performance mode.

Cheers, Hans.

1 Like