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

Just to be on the safe-side, I’m going to replace the i3 in my Nucleus with six hamsters going around in wheels (in parallel of course!) to do the processing.
Let’s face it, they’re bound to contribute far less noise to the network than a piece of silicon, and if I want to underclock them, I can simply tie their hind legs together, for even less noise :nerd_face:


Well, it depends on how much bathwater there is. It’s Great Lake’s size now. I would say there is a lot of information there. Knowledge? Sure…but finding it is like searching for a needle in a hay stack.

1 Like

There was credibility to begin with?


Sounds to me that you imply that the Roon architecture as it is today did not came to be without testing, tinkering or any listening to possible improvements. Personally i would find it very wrong if Roon dev team would not explore avenues for improvement.


I agree but that wasn’t the topic of discussion in this case

What was the topic then? Question asked in ernest by the way.

To summarize, the question started with the title of the thread. Roon basically said “no”, and these are various reasons why that couldn’t be implemented in a good way considering all the balls it has to juggle (multiple users, etc).
After that it devolved into the inevitable “you can’t hear a difference anyway”, and here are the reasons why.
(or at least my translation).

Thanks Scott, sort of my conclusion also after extensive re-reading and trying to understand this thread.
One question i still have: As Roon stated they cache in a steady stream of data, and the dac is fed from the cache. If understood correctly, is that not also " from memory" ? And if not, what is the key difference w.r.t. audio?

I’m not qualified to address this one. Over my head, tech-wise.

The (few) people who are asking for memory playback want the whole track cached in memory before playback begins. They believe that the (small and intermittent) amount of processor activity involved in keeping a smaller buffer from going empty is detrimental to the sound. Some people who make expensive streamers also propagate this view. If you want to test your system and see whether processor activity upsets it, try loading a big spreadsheet while playing some music. Does that make an audible difference? Other players like J River Media Center have a memory playback option that got put in to appease a small number of a particular kind of audiophile. Didn’t make any difference when I tried it. And since you would need an awful lot of memory to load a gapless album, it is just as well.


In theory it may. I have a nuc7i7dnke with rock. A Qutest connected via USB. Have not yet noticed any difference in SQ when f.i. was adding an album to the catalog etc.
So i am happy.
Having the SMPS wallwart plugged in with the right polarity did make a difference. As for any other equipment in the chain.
As did earthing the core switch in my network.
But this is not in scope of this thread.

1 Like

A Qutest is galvanically isolated so it is hard to see how it would be susceptible to noise on the USB input, even if there was any caused by the cpu buffering. Though it is also the case that Rob Watts himself recommends running a laptop from batteries for the absolute last ounce of sound quality.

1 Like

About 800 MB max for a 16/44 PCM album, Hi-Res, quite a bit more, DSD… more still… I’ve got 16GB memory in my Mac Mini, so… it’s do-able, but… still…

Is it necessary…?

I’m not getting into that argument… :slightly_smiling_face:

… but if the argument is that cpu noise upsets the audio, it’s not as though buffering a whole track would mean the cpu is off while music is playing. It will be working parcelling up all those tiny little USB packets, sending them somewhere, responding to interrupts, checking the buffer and who knows what else, so it is likely to be upsetting the audio in any case. Best look for hardware that isn’t so flakey, imo.


I agree, system is always doing something. So the question then is something like, What hardware will give the least disturbance of the audio stream?
With disturbance i mean anything that gets added to the data by the system, and that should not be added. Such as noise, disturbances in time etc.
Also sounds to me as something that can be supported with the appropriate choice of measurements, in addition to listening observations.

Roon’s recommendations on this are here.

Given that Roon’s recommendation is to separate the core/server from the endpoint via an ethernet network it is also worth a read about Roon’s network protocol RAAT.


Well, i use a nuc 7i7dnke with rock, as core and endpoint with qutest via USB. Would a separate endpoint, via ethernet connected, still provide a benefit?
My nuc runs almost idle.
Find this hard to believe, that the extra complexity to have RAAT do its work over an extra string of connectiion elements would give less noisy music reproduction.

So what is it that an separate endpoint cause the suggested beter SQ?
That is what i would like to understand.

Computers are inherently electrically noisy things and plugging one into sensitive DAC could cause the noise to get in to the analogue domain of the DAC. Decoupling via ethernet to connect to a separate playback device breaks this as ethernet is galvanically isolated. However then you get in to the old ethernet also has its own variation of noise that can also get into the DACS analogue circuitry. In the end use what you think is best or have the means to try or you end up going down the rabbit hole their is no escape from.

The separate endpoint is also a computer, thus noisy.
So i do not understand.

It is a very small computer with a minimal OS and no other software on it, so the noise it generates is minimal.
This is the theory.

But a NUC/ROCK is pretty minimal too.
And a modern DAC like your Qutest is not very sensitive.

I have a Nucleus and a Hugo 2, I have tried it direct and with a MicroRendu, and found no difference.
I’m not saying the MicroRendu doesn’t work, I’m saying it’s purpose is to remove a problem, and if you don’t have the problem you don’t need it.

There is a lot of theorizing, people arguing for such separation.
There are some who measure and find no difference.
Some argue you should try it and decide by listening, but then you have already paid…

I think there is a belts-and-suspenders problem. Even if you accept that electrical noise is a real problem.
Belt: get a modern, minimal computer like the NUC/ROCK instead of a big Windows desktop, and get a modern DAC.
Suspenders: get an isolation device.
I found I don’t need both.