Core as endpoint vs dedicate endpoint

I am constantly fiddling around with my setup and find that using my dedicated 5 year old I7 16gb desktop as Core and endpoint for my Mojo DAC to be a far more stable solution than if i connect my laptop to core via ethernet+router, and use the lap as endpoint with the Mojo.

Problem is, I use the laptop for work and loading pages and stuff causes the Mojo to stutter on occasion (enough for it to be irritating).

I am wondering therefore what are the sonic benefits of using a separate endpoint connected to core? Or is it really to avoiding glitching/overloaded resources on the core (notwithstanding of course the audible benefits of a fanless setup remote to the core)?

What are your findings? Better imaging? Layering? Detail? etc?

For me, I don’t really notice anything sonically between the 2, but I rarely get the chance to really listen critically :slight_smile:

Why not try something cheap and cheerful RPi based maybe or Allo sparky or microRendu perhaps depending on your budget

Seperating to a dedicated core and separate endpoints is the best thing sonically for me. Building a solid core machine that does nothing else has improved things even more. I run a small fanless pc with ROCK and it’s the best thing I have done to the system by far. I can’t over emphasise how much going to a fanless dedicated core improved things, it blew me away and I am not one that normally believes it should make a difference. Then get a solid endpoint and use your laptop and not worry about music. Yes its more expensive and has more kit but for me its made a huge difference.

1 Like

I have a Allo usbridge (sparky). Before having Roon installed, I moved from a RPI to the sparky solution which gave me a better SQ on my Chord Qutest DAC (subjective opinion).
Since I have installed Roon, I now have the possibility to connect the core directly to the DAC and test if there is any gain of adding a dedicated endpoint ( the allo usbridge). Before reading your post, I was already thinking of doing this test. So I can let you know if interested? I will do that in the coming days.
A part from sound quality, It will be interesting to check app reactivity if used as an endpoint depending on PC/NUC specifications and number of endpoints to manage in parallel (especially if some flow are upsampled, e.g. DSD256) .

1 Like

thanks Wiz, but it was not really my question. Im ok using my core to connect my DAC to, just wondering on the mileage of others. But your options are interesting to hear about all the same :slight_smile:

great, would love to hear. The Qutest looks like a great DAC. I was checking it out the other day, as my oldish mojo is playing up a bit with charging and stuff. Bit out of my budget though :slight_smile:

it would be nice to hear about sonic benefits Simon, if possible?

PS…my desktop is pretty dedicated to Roon. I have my music HDD’s connected and it does pretty much nothing else, until my son wants to play Minecraft or ROTMG on it :wink:

@Sallah_48 sonic changes or benefits as so subjective that i try to say out that side of things. There are plenty of things to try that are cheap and or affordable where you should trust your own ears not the ears of others as everyone’s setups are so unique.


To me the move to a fanless pc improved things the most. Its hard to describe but it made it what many refer to as a veil being lifted that I did not know was there I was happy before . Everything just feels cleaner, less annoying high frequencies, more defined bass and overall less fatiguing. As I said I am one that normally dismisses all this as audiophoolery and went in not expecting any difference it was more of a way to get the unit out of the cupboard as maintaining its temp was hard in their. And it’s been a revelation. Both my systems sound so much better for it. I had no expectation bias as I expected nothing to change. With it being now silent and fanless it could be used as core.ans endpoint but I don’t have an external DAC so to me its not needed.

Are your files stored locally on your core machine or are you pulling from a NAS?


On HDD’s attached to Core via USB 3.0 - 4tb and 8tb devices

1 Like

Hi Sallah!

To my understanding, the reason you want to use a separate endpoint is because it should be doing NOTHING else. The concern with a busy PC is that processing and other mobo activities related to decoding FLAC files, operating storage, getting email, whatever, throw out electrical interference that could affect the way audio streams through USB to the DAC.

So you have another endpoint that isn’t doing anything but streaming the audio from the network out to the DAC.

When you actually use your laptop, even for something other than decoding FLAC, you’re defeating the purpose of the separate endpoint.

I also doubt that any laptop is ideal as an endpoint. They’re not built to minimize electrical noise and everything is so close together, hot, etc.

In my experience a quiet endpoint DOES make a difference. I went to the level of buying a SOtM USB adapter with some big ole’ capacitors to filter out that noise, and have a fanless CPU, SSD, nothing moving, and the background is definitely more black. For me it creates the perception of added detail. Except when I turn on the TV…ugh.

I’m not a believer in one-way fuses, cable risers, little pebbles and such. But I do think that standard PCs are not ideal endpoints and should be minimized in all possible ways. Check out the computer audiophile builds, especially the few earliest ones. Those are good audio PCs to use as endpoints - probably overkill, but it’s just a straightforward sensible way to optimize PC hardware for audio.

1 Like

I ran my main stereo output as optical digital output from the headphone jack of the same ancient Mac mini I was running the core on for a couple of months, into an optical input of my stereo receiver, which has an internal TI PCM1754 DAC, with no issues. So I guess my experience was the same as yours.

I eventually moved the core to a new machine, retired the Mac, spent far more than was necessary on a new external DAC, and replaced that endpoint with a simple RPi running RoPieee and connected via the Pi’s WiFi to the network, and via the Pi’s USB to the DAC. Again, no issues except for the rate-switching bug.

1 Like

thanks James.

Makes sense. I went to the trouble of Routering up the house and cabling ethernet everywhere, lol. Next stop - hardware refinements.

Hi Sallah

I had a similar setup to yours: i7 7700K based machine with 16GB ram and local SSD storage for my files that acted as core and endpoint. I then tried using my pc as the core, and my macbook as an endpoint. The SQ went up. I turns out my pc generates more electrical noise compared to the macbook and it’s very audible (lower noise floor, better separation), on both dac’s I use (Dragonfly Red and Audioengine D1, both don’t have jitter reducing circuits higher end dac’s might have). Since my PC has audible fan noise ans coil whine (even with a really high end custom watercooling) I wanted to be able to shut that down, so I bought a i5 NUC and put the core on there…

There was no audible inprovement through the speakers, but less although still present fan noise of the ROCK. The SQ from the ROCK as endpoint or the Macbook pro seemed similar so I ended up using the ROCK as core + endpoint, until the fans bothered me too much again…

Then I moved the ROCK to the server closet, and kept using my macbook pro as an endpoint. The SQ was really good and my listening environment was without fan noise, but the setup was rather inconvenient, with dongles and the fact that the macbook is primarely a workhorse that I need to take everywhere…

So, I hooked up an old iPhone 5S to a usb 3.0 CCK with the Dragonfly… This is the best setup yet. SQ is at least equal to the MBP and 100% silent in terms of mechanical fans / HDD or coil whine… I guess that for these DAC’s this is fine…

My advise: remove everything with fans away from your listening environment and get a dedicated endpoint that is either battery-powered or has a proper power design suited for audio playback.

This doesn’t really make sense. Since Roon does the decode on the server side, your end point just takes RAAT packets and sends them over USB. This takes basically no processing power, because the end point does none of the work. Therefore, the core is where it’s important to have nothing else running / linear power supply.

You may also want a linear power supply on your end point because it may introduce noise into the USB (maybe), but any decent audio hardware shouldn’t be affected by the minimal amount of noise there may be.

This is why I have my set up as the core & end point on the same machine, which is headless, driveless, fanless, and has a linear power supply. This dedicated machine does nothing else, whereas my file server (which would be running Roon Core otherwise) is doing all kinds of stuff in the background and has a switching power supply.

I’m very happy with my configuration and it sounds great. Better than having the Core on my fileserver.

You can also do a dedicated Core machine w/ linear power supply and dedicated endpoint w/ linear power supply, but that just seems like overkill and nearly doubles the cost (good power supplies are not cheap).

There are no sonic benefits, and a Core i7 w/ 16GB of RAM is more processing power than Roon could ever conceivably hope to use.

I’d like to add that everyone hears differently. Try what you can afford :smiley:

I think some folks are not quite getting the point: high processing power on the core is what you want, but that is EXACTLY what has the potential to pollute the audio stream if also used as the endpoint. High wattage CPUs are part of the problem with electrical noise, and they also produce heat, which requires fans, etc. It’s best for that device to be sonically separated from the listening area and not to be the endpoint - have the interference stop at the ethernet port.

Of course what I said makes sense. It is the “core” of Roon’s endpoint-based architecture.

If users are happy with a combined core and endpoint, that is fine, I am glad you are enjoying it. But the clear point is to have exactly as described here:

Yes, exactly, and that “no work” is what keeps the audio stream through USB as clean as it could be. Using the core as the endpoint, even one with no moving parts, potentially adds electrical noise. If you’re not experiencing that, great, it just means you like your setup.

But theoretically, the best thing is to have the endpoint be dumb, do nothing other than stream from ethernet to USB. And building the cleanest, least electrical device possible as the endpoint is the goal with that philosophy. It may be overkill, but the logic is consistent.


The linear power supply for the digital side also seems unnecessary. Archimago likes to point out that he doesn’t use linear power supplies, and can’t find electrical noise in his measurements. And, as you point out, modern DACs and receivers incorporate isolation technology to make sure electrical noise on the USB bus won’t get into the analog side, anyway.

1 Like