A server is a completely different entity. It is supposed not to have a "sound" but it has one. They all sound different. No one knows why they sound they like they do - "experts" disagree every minute on them. There are too many ways of connecting servers to DACs.
We have some idea of why they sound like they do. It's mostly the jitter of the Master Clock, wherever that resides. With USB however, it is clear that different player software and even different revision levels of the same software change the sound. It is clear that different USB ports and how the logic is powered changes the sound. It is clear that different USB cables change the sound. These are some of the reasons why I'm focusing more in Ethernet now and less on USB. Optimization of USB is not easy, so the current way to get there is to buy a server like Antipodes or Aurender N10 which take care of this for you. You can also eliminate some of the sound dependencies of USB by using an Ethernet to USB converter from Sonore. Be aware that even buying the best server for USB will not fix the jitter in a poorly implemented USB interface in a DAC.
There are advantages and disadvantages of each connection method: USB, Ethernet, I2S, S/PDIF and AES/EBU.
Async USB - convenient, but too many things to optimize and too many variables that affect SQ. Can be galvanically isolated. Dependent on the Master Clock in the interface or DAC, not the server. Usually requires new drivers for PC for each OS release.
I2S - this is a great connection because it connects directly to the D/A, however the Master Clock in the server had better be really good. No way to galvanically isolate. Good I2S cables are expensive.
S/PDIF - this can be extremely good, depending on the S/PDIF receiver used in the DAC. Dependent on the Master Clock quality in the server. Can be galvanically isolated with a pulse transformer.
AES/EBU - has the advantage because it is differential of having some common-mode noise rejection, but can also be galvanically isolated with a pulse transformer. Disadvantage of requiring higher voltage, so another stage of buffering adds more jitter. Dependent on the Master Clock quality in the server. Difficult to get a good cable impedance match due to the XLR connectors.
Ethernet - Very few variables that affect SQ. Can use inexpensive cabling. Shares the LAN, so traffic can interrupt playback. Automatically galvanically isolated, however some steps required in power supplies for router or switch to make it ultra-low noise. Dependent on the Master Clock in the interface or DAC. No reliance on special drivers. This is the easiest route to good SQ IME.
The perceived value is another reason - why paying ten times the price of a top desktop for a machine that is unsellable and we are not sure that we will like? As always, YMMV, I would appreciate to read different opinions on the subject while assembling my server ...
I recommend to not spend big bux on a server. Instead, get a good Ethernet interface externally or in the DAC and just use any decent computer to run Jriver or Roon. You can still stream Tidal or play your rips etc.. Rip CD's with XLD on Mac and rip with dBpoweramp on PC.
Steve N
Empirical Audio