There are a lot of conversations and ‘wires’ about power supplies on ‘ROON’ NUCs or other types of tweaks for ROON servers. Now we already use ROON Rock on a Gigabyte Brix NUC (Core i5 model). And we’re very pleased. Response is fast – even with many zones – and playback is great. Since we are also curious if tweaks do something there, we installed ROON in a VM (Virtual Machine) and compared it to our Brix which we provided with a Sbooster linear power supply for this test. Is there a difference?
No. There’s no difference.
And now the explanation
The set-up of these two machines is almost the same: ROON Rock is in fact a stripped down linux version of ROON. We have therefore installed Ubuntu on our Unraid server (RAID array, SSD cache, 10-core Xeon (E-2690V2) with 48GB RAM) in a Virtual Machine. We have assigned 2 cores / 4 threads and 8GB RAM to this Ubuntu installation. This is roughly comparable to the Core i5 of the NUC. In this Ubuntu-install we then installed the Linux version of ROON.
Now it is possible to switch continuously between these two installations, but we wanted to do it easier: we have two Metrum Acoustic Ambres which we connect to the Pavane-dac via coax. This allows us to quickly switch between these two servers. Just start the same track and change inputs. It doesn’t get any easier.
In fact, the ROON server does little more than process a data stream of (music) data and send it to an endpoint. This is done via the so-called TCP protocol – Transmission Control Protocol. This is a protocol (part of the ip-stack) that knows correction. It’s a so-called liaison protocol. This in contrast to UDP (without correction layers). UDP has basically no error correction. TCP does. In other words, TCP does not allow errors to occur in the data.
Of course, this does not have to mean that there are no noticeable differences. After all, with network cable we also hear differences. This is due to the electrical connection you make between your network and the streamer. But, uh… this problem doesn’t really play with a ROON server either. After all, it runs on a PC, NUC, NAS… That’s the polluting element.
What could be left is jitter. But we can exclude this almost immediately. Because streaming – also at ROON – is packet based. In the header of the data packet is where it comes from, where it should go, where it belongs in the data stream (the ID / datagram, supported with the flag and the fragment offset). Jitter as we know it with, for example, spdif plays no role here, because the receiving side simply receives the data packets and glues them together in the buffer, turning them into a music stream during decoding. As long as this is done properly, there can be no difference between source A and source B.
Is processor power then a determining factor? It could be. An undersized platform could make mistakes. But then this will manifest itself in hick-ups, outages, very slow response. Not so much in timbre or imaging. Editor in question runs private ROON on a Synology NAS (DS1517+ with 8GB of RAM and SSD-Cache) and that runs very well with two or three zones active. And we can’t really call a quad core Intel Atom a powerhouse. Isn’t that right?
NB: there ís an audible difference between Roon on a NAS or Roon on a NUC! A Nas-system does sound less ‘free’ than a Nuc. We heard this in later tests we did at home and at the office.
Whether you run ROON on a NAS, a NUC, an expensive desktop, in a Virtual Machine … it doesn’t matter in sound terms. We really haven’t noticed any differences. So forget about those expensive tweaks on the server side of ROON… Unless you plug a NUC into the same connector block of your hi-fi system (sometimes you can’t avoid it), it really doesn’t matter whether you connect the laptop adapter or a linear power supply to it. Make sure you have a decent network connection and you really are the ‘man’ (or ‘female’) as we Dutch say.