Ok, here’s an excerpt from one of the posts from OP Tobes:
Firstly, let me say that I find it baffling that separating these functions can have an effect. I only do pcm upsampling and it doesn't use that much cpu grunt.
An M1 Mini running Roon and HQPlayer with pcm upsampling would typically be using around 5% CPU resources - though a sizeable chunk of memory is in play.
I'm sure there are plenty of people running Roon+HQP on an M1 who are perfectly happy - I think it sounds fine.
That said when comparing Roon+HQP to HQP standalone (with its own minimalist player) some significant differences are apparent. At least to my ears, in my setup.
HQP without Roon sounds more transparent. The depth and space around sound sources is clearer and there is more focus and texture/timbre to sounds which makes music sound more realistic and present.
So, IME, HQP sounded better standalone than with Roon - this was true even when I separated HQP onto an M1 Mini and ran Roonserver on the LPS powered i7 Mini (both connected to the Etherregen network switch) and the UltraRendu acting as the endpoint to my usb DAC.
So it was a surprise to find that Roonserver running on a stock M1 mini with HQP running on a second M1 mini, along with the Roon GUI, didn't show the same deficit. I'm not sure why since resources wasn't an issue with i7 Mini.
Edit - and his conclusion from another post:
For perspective, the change from i7 Mini Roonserver to the M1 Mini Roonserver was much bigger and more obvious than the addition of the EtherRegen. But would it have been as obvious if I didn't have the EtherRegen in place first? Hard to say.