Best Linux distro for building a Roon Endpoint

I took statistics and I’m not talking about network latency. Did you read Andy’s response?

How do you know the processor’s buffers aren’t too small? How do you know whether interrupts aren’t impacting the processor? How do you know the drivers don’t impact access to the processor or maybe the buffers are too big? How do you know latency doesn’t impact sound quality?

OK. How would we be able to measure the impact an OS has on performance?

Your ears combined with input from others who hear similar benefits. Until someone with the right tools can conduct a scientific study and publish them.

You can use a Pi with a HAT that provides an SPD/IF output like the Hi Fi Berri Digi, or the equivalent IQ Audio (can’t remember the model that provides digital out).

I don’t see any advantage of using a full blown PC for just an endpoint, especially not one with Windows running on it, with lots of irrelevant processes going on in the background. That would probably apply almost as well to a full blown Linux desktop distro. ROCK would be overkill, if you’re just using it for an endpoint. If you want to run your core on the same box, different story.

1 Like

I just ordered a Canakit Pi 3 B+ to try as an endpoint. No hat. Just use the USB from the Pi out to my Mytek DAC, like this guy does. $61 including 32GB card.

Well, you could try a test setup like this one. Use different OSs on the Pi and see if the results change.

But is it possible?
To run ROCK as “only” roon endpoint?
I have a NUC with a Pentium N3700 with a very low TPD of 6 Watts waitung here to try.

Yes, it should be possible.

1 Like

Yes, it IS possible…

1 Like

Thanks Geoff and Daniel :sunglasses:

The core is on another Computer.

Will the lightweight ROCK also be able to use WiFi?

Well, you can certainly set it up (using the ROCK web interface). However, when you do, you will also see a warning:

Ethernet is more reliable for audio transmission, WiFi should only be used for control.

1 Like

I have been using (and loving) ROCK as endpoint software for a while on a single core Atom NUC.

1 Like

hi there I am tryng to understand if sound is affected or not if latency is not controlled by any method or hardware . In my case I use as endpoint an old pc with a amd processor and 4 gb of ram. Do you think that working on lowering the clock values on processor as well as ram memory should help in any way to improove the final sound out of the box?
any help here should be very appreciated. amd

I am running Roonbridge on the endpoint machine connected to the dac via a usb to spdif halide bridge .

Your biggest gains would be in dealing with power.

1 Like

I really don’t know, but I’d doubt it. It’s hard to come up with a reasonable hypothesis as to how that would happen. I’d be inclined to believe what @Henry_McLeod says: what you’d mainly notice is reduced power usage. But only by a tiny amount, I’d think. Perhaps the equipment would run cooler, so less fan noise?

On the other hand, there are always software/firmware bugs in modern downstream devices, as with this XMOS chipset problem in a number of DACs. Many of them are timing issues, race conditions in various parts of the software. So changing the upstream timing would tend to affect those downstream issues, which may shift the balance either towards or away from better SQ – almost impossible to tell in advance. I’d think the most hopeful thing to do in such a situation is continue to pester your vendor to fix their firmware, and stay up-to-date on firmware upgrades. What a pain!

Or, you know, drop digital and do everything in a firmware-free analog world, back to vinyl and tape and tube amps.

This is not my experience. I use a server core (NUC) and endpoint output device (microRendu 1.4), and have found changes at the server level (power, OS, etc.) to still have a big impact on sound quality.

I can’t explain it, considering the RAAT architecture. Seems like the server noise shouldn’t matter one bit, but it definitely does.

If it isn’t measurable, it’s likely psychological.

(and no, I’m not saying this to be dismissive with your findings, just to stress that expectation bias is a hell of a drug)

Or ground loop.

1 Like