There are three issues at hand. Quality of usb out of server/streamer, combining processes within one box and separating server from streamer, aka two computer solution.
Many servers don't optimize usb out, take usb directly out from motherboard, this is subject to all the noise created from the many processes on motherboard. Now, some servers optimize usb out such as the Mojo with JCAT card or some may use Pink Faun, that takes care of that issue. As a matter of fact, I was contemplating the Mojo server as possible candidate for eventual replacement of my present server until Benjamin notified me he will no longer be manufacturing.
Next issue is single computer or server/streamer solution. In this case many more processes taking place within singular component, more noise means less resolution. Dividing up the processes between two components, separate server and streamer means less noise generated within each component. While that may be of some benefit, the greater benefit as I see it is the ability to optimize your network post server. With separate server one can do FMC or Eno post server where it does most good. Prior to server isn't treating noise generated within server and or server/streamer all in one unit.
With Opticalrendu one gets FMC, quality streamer and usb renderer all in one package. Noise eliminated with FMC, many choices of what to stream and usb optimization. One Femto clock is for network side, other Femto clock for usb, proprietary circuit board optimized for low noise.
As previously mentioned I've also modified my server for direct ethernet out which means no more detour back through another ethernet cable and switch. More direct path means less chance for noise to enter chain.