The author notes that DSP functions demand additional processing power and recommends isolating the server from the rest of the audio chain (I assume he is using Roon).
What’s missing, however, is any measurement or evidence of how much noise is being introduced. Extra processing may generate noise, but that doesn’t automatically make it audible. Interestingly, the author himself relies on a Raspberry Pi as a streamer—clearly, he does not consider that to be a significant noise source.
I have a setup similar to his (Roon running on a server). I also use a stand-alone streamer with Qobuz Connect, and frankly, I cannot detect any audible difference when switching between the two.