CD transport versus music server:
The first and perhaps most obvious variance is in the masters, version or issue of the source medium. If I have a CD recording I want to compare I have to rip that particular CD first, as online streaming versions or digital downloads often sound different, they can sound similar, better or worse, the variation in this is unfortunately not subtle, but different CD issues of the same recording can sound different too.
Further to this I have to rip it to uncompressed .wav to the Extreme's local storage if the goal is to achieve a similar sound signature.Note this is not a declaration of superiority for the uncompressed .wav format as there are many factors in play here. Compressed flac files are smaller in absolute size which turns it into a file transfer versus decompression game with differing noise signatures depending on the construction / configuration of the server's hard- and soft-ware. With that said, uncompressed .wav from local storage sounds closest to a CD transport.
Then we have the playback software used, the very first version of TAS using Logitech Media Server as it's front end and squeezelite as it's backend, the second version of TAS using Logitech Media Server as it's front end, a squeezelite to ASIO bridge and our own developed ASIO player (XDPlayer) as it's back end. We now have XDMS being fully coded from the ground up and of course there's Roon. Then with Roon you also have the option to replace the endpoint software with Squeezelite or HQPlayer, which are options most server vendors offer nowadays. Of all of these the current XDMS alpha does sound closest to a CD transport but the upcoming XDMS "NSM" deviates by being "smoother" /less "choppy" then CD playback, even with the same 16/44.1 source file.
Another significant contributor is the interface used. Perhaps unsurprisingly SPDIF and AES/EBU sound closer to CD then USB/Ethernet, simply because you use the same interface. USB can be better but is subject to its implementation in both Server and DAC, IMHO to a much larger degree then AES/EBU / SPDIF.
The biggest "elephant in the room" however is the sheer amount of variables introduced in streaming / server based solutions. With a CD transport variables are largely limited to the AES/EBU or SPDIF interconnect, power cable (and power quality / filtering etc) and vibration control, aftermarket footers, racks, flooring etc. With servers this extends to your network, Wi-Fi and all devices connected to this. For example, especially when using Roon which is the most "sensitive" playback software to this particular variable, Image placement can vary, change positioning and/or diffuse because of network activity. This led us to design a full network setup solution including the recently released network card and switch, and the router with Wi-Fi shipping soon. With these in place imaging is rock solid and purity/clarity is significantly enhanced.
With all of this mentioned, I do have superior performance here over CD, assuming the same source file is used, mainly in the areas of sound staging (size), detail retrieval and instrument decay, and yes, wile using Roon. Without the Router/Switch you will need XDMS for this though.
A small addition, IMHO the upcoming upgrades, like the TACDA / TACDD interface and BPS elevate everything to previously uncharted territory, I'm aware this is a bold statement, but to my, and meanwhile a few other ears, perhaps not that bold, time and your impressions will tell.