Does Roon download entire track into RAM? [Memory Playback Discussion]

I’m a little confused though, because data is obviously cached pretty much at every stage in the chain, from basic disk i/o, output buffer, tcp/ip frame buffers, input buffer, isn’t it effectively being played back from RAM anyway? Something else must be at play if you hear a difference. And I will say, as skeptical as I am, I did hear a difference when Auralic (to great fanfare) introduced RAM playback as a new feature. Roon definitely need to stay in the game here.

1 Like

Wouldn’t it be more useful to have a dialog rather than an argument on this topic? I mean both sides. From one side we seem to have a data point that a single device sounds better with RAM playback, and from the other simple denials. This is useless.

Lumin U1 Mini as endpoint here :slight_smile:

as it also works as Audirvana endpoint, I just gave the latest Mac 3.5 version a try (btw: ugliest and most cumbersome UI ever that took over one year to be released :roll_eyes:). sounds slightly better than Roon, though (with everything else in the chain being the very same)

no idea why and not going back to Audirvana but… :wink:
(and signing off this thread :stuck_out_tongue: )

Personally, I don’t care about adding this functionality. Especially so if it would compromise other aspects of Roon I love, like concurrent streaming to multiple endpoints and ability to hop around quickly in an album.

I don’t particularly care because my listening experience has been that the difference sound-wise among variants of the computer front-end generally has been less than overwhelming. I’ve tried JPlay variations, JRiver variations, different USB settings, AudioOptimizer variants on Windows 10 Pro and Windows Server 2012. Sure, there were some differences here and there, and sometimes they were rising to the level of significant. With some of these (Audiophile Optimizer on Windows Server 2012) there was a meaningful difference, but switching to ROCK on a Nuc streaming to a RoonBridge gave pretty similar performance.

Now, under my conditions (C.A.P.S.V3 Zuma) and an MSB Analog DAC, there might not have been the right set of conditions to allow pronounced differences. Maybe if you had a noisier computer and a very resolving DAC the differences would be meaningful. But that’s not my experience. I feel like I’ve goofed around enough going down this path that I’m no longer interested in spending time chasing these differences anymore.

I won’t say that Nothing matters here. DSP convolution filters are amazing. Can’t live without that. I’m just saying my expectations for playback from RAM are low.

A dialogue is a waste of time. The issue can only progress with evidence. Those who believe that memory playback makes an audible difference need to demonstrate their ability to hear that difference under acceptable, double-blind conditions. If they can, then an engineer has something to work with. If they can’t, then it is perfectly reasonable for the sceptics to say there is no difference, and there is nothing to work with. Roon cannot be held hostage to the unproven whims of a handful of anonymous forum posters.

Well, to be fair to some posters, I don’t think we have a hostage-taking situation here. :wink:
There was a question of how this works in Roon. Roon answered why they don’t do it based on how they want the software to work. Posters expressed their interest in this for their reasons, and then everything snowballed into a mess.
But I do agree further dialogue at this point is a waste of time. Everybody has made sufficient points and attacking each other doesn’t get us anywhere.


Yet another one yelling “prove it to me”! As much I refrain to engage with such crowd, I often cannot help it.

As I said it earlier, I hate this is happening, but Roon with SqueezeLite experimental feature on my Innuos ZENith MK3 sounds better than RAAT. I cannot explain it why, but I experience it every single time I compare. It’s very noticeable. And easy to switch and A/B with a press of a button on my iPad Pro, it takes less than ten seconds.

And I have no way to prove it to you. Unless you are in DC area, in which case welcome to my place to hear for yourself. I simply share my experience. That’s the extent of it


The point is, you’re probably in the minority.
The vast majority of Roon users like things just the way they are, and like the way the software functions, and the way Roon sounds.
Why should Roon adapt to satisfy a small minority who ‘believe’ something that isn’t proven, and therefore remains speculation?
Obviously, if you don’t like the way Roon sounds, you have other alternatives.

That’s correct! Except for the “Majority “, including yourself, have never tried the alternative

I, and many Bryston BDP owners, hear the same thing when comparing RAAT and MPD. I can use them both networked playing the same file. MPD sounds better. I thinks its as simple as RAAT being heavier on the CPU which results in more noise generated within the endpoint. MPD probably has less processing involved.

This was Brian’s post about BDP-1 and how RAAT and MPD are handled: Huge number of lost PCM samples in network endpoint output [ALSA software device]

Is that the case with other endpoints as well?

And frankly, I can’t be arsed about trying any ‘alternatives’. RAAT works perfectly for me, thank-you.
Why would I want to ‘try’ alternatives when what I have with Roon/Nucleus/Lumin A1 just ‘works’. And perfectly.
As I said, it would seem that you’re clearly in the minority…

Yes, I am in minority. For now. That does not change my experience that I am sharing. What’s your point? Does your LUMIN do memory RAM playback? Certainly no. I should know, I used to have the T1. So, you are basically just speaking theoretically, like an armchair quarterback.

Oh… and I love my Roon. I am playing Roon on RAM memory right now!

1 Like

If your Innuos sounds better with squeezelite than with Innuos own software, then it is Innuos you should be inviting round to your house to hear it, and Innuos you should be asking to fix the problem with their hardware and software.

You seem to think that if folk aren’t using Innuos they must be using Chromecast or Sonos. If not, that the best you can do…

Really? That is not what it says here:

There it says that the Innuos box can function as a Roon Core or Roon Endpoint (player). That would mean no SqeezeLite…but hey, you know everything…


I listen via Meridian DSP SE, it sounds amazing (Jaw dropping) why should I bugger about when what I have is soooooo great now?

1 Like

I was replying to the other guy or girl. Please read what he said first. That pertained to using Innuos as a streamer, and using own streaming “program” to stream. One can surely use Innuos as a Server only, and use another device as a streamer.

Anyways, this is clearly becoming a moot point here, as the other folks have mentioned, with the mob / herd mentality in full display. You boys can take it from here. It’s all yours. Over and out.

1 Like

Roon have already said that full ram playback brings no advantages to RAAT only disadvantages so it’s not going to happen so we all need to move along after all they know their product better than anyone.

Ram playback might well help squeezelite elevate above RAAT, I can’t say if it does as I don’t own InnuOS or used a Squeezelite endpoint since switching to using Roon Bridges but as so many other variables can change perceived sq then I will not judge until I have experienced it myself. But it does not mean it would help RAAT, Brian already has said as much.

Whether RAAT is your preferred choice from a sq perspective or some other protocol is always up for debate, it comes up a lot here and other forums, especially from users on the high end ki. Maybe RAAT still needs work in this area without RAM playback.

Like everything in this crazy hobby we like to assume things do or don’t change anything. I was sceptical about using linear PSU for my ROCK but intrigued by so many saying that it improves SQ. I bought one and have noticed very little if any difference. Same went for internet cables, I recently decided to try one of the fancy ones, again made no discernable difference. I recently changed my extension cable to a Russ Andrews one to see if that would change anything, to my surprise it did and it’s stayed in place. I am not going to buy an InnuOS to try out ram playback but I will keep open minded that it could be better, not that I have an issue with how RAAT sounds. But then you don’t know what you can be missing until you sometimes try or we would all still be using crappy bell wire for speaker cable.

Those that want RAM playback can still use it without any extra dev from Roon so let’s all just move away from the hostilities.

Intel CPUs are based on a die of 14nm and AMD is working on a die size of 7nm. That’s a lot of electrons, moving at the speed of light, in a very small space.

If what you believe were true, computers wouldn’t work at all. Playing music would be the least concern.


It’s not what I believe. I’m not trying to ‘believe’ anything, but rather just understand and make sense of my observations. If there is a better explanation for it, I’m all ears.

Is the power supply draw/fluctuation footprint in the endpoint different (worse/between) between RAAT and say DLNA or MPD in various devices? Are there more number of processes happening between different protocols. In the above post, Brian does provide how MPD and RAAT differ in workload.

For example, Linn and Naim have both measured differences in power supply rails between WAV and FLAC playback. Linn says they have filtered it down to not make a difference in the outputs. Whereas Naim still continues to say WAV is preferable for best SQ. (This isn’t about FLAC vs. WAV, but if they can detect differences between WAV and FLAC, then it’s not impossible that there aren’t measurable differences between RAAT and whatever else.)

The mechanisms in these cases exist and do have merit. Whether they have any significant measurable and/or audible differences, for their personal use everyone can decide that for themselves by various means. Read the existing threads, published measurements, and/or listen for yourself. If you think none of this matters, awesome. No need to worry about it. If you do hear differences, is it enough to stop using Roon and all its benefits? To each their own.

FWIW, I love Roon and its responsiveness. I’m not advocating for memory playback if it will mess with the current responsiveness. It works very fast with all my devices. Does it sound the best on my favoured endpoint? No, and that’s fine. I’ll continue to use MPD with the BDP-1 as needed if I desire the best SQ. If I was interested in getting the best SQ from Roon, I’d definitely audition and consider a recent endpoint that doesn’t let these (however tiny) differences in processing affect SQ.

Ultimately, I think it’s the manufacturer’s responsibility and not Roon’s to get it right and have it sound its best with Roon if that’s the goal. Roon just has to make sure the experience is stable, trouble free, and bit-perfect.

I don’t hold it against the BDP-1 that it doesn’t sound as good with RAAT as it does with MPD since it came out almost a decade ago and wasn’t designed with any of those things in mind. It was primarily designed to sound good with straight USB playback. I didn’t buy it with Roon in mind. Later on they added Roon support, for which I’m happy.

With Roon/RAAT having been out, if I was getting a new endpoint with Roon as an absolute must and as the primary playback method, I’d make sure that the device sounds acceptable with Roon for my standards.

1 Like