Hi Romaz,
My approach has been entirely different. Instead of using HPCPUs to correct incoming noise while adding their own noise to the outgoing signal I get rid of the incoming noise before it reaches the server. This has several benefits. 1. It means the CPU load is lower. 2. It means that a LPCPU can be used so 3. I get less incoming and CPU-generated noise going to the DAC and less latency throughout.
It appears you have curated a system for yourself that has made you very happy and so that's what matters. The engagement that your system brings to you that you described in your previous post is what we all aspire to and so if you have indeed arrived, then congratulations. My wife and I attend the symphony once a month on average. If we are fortunate, on a good month, we'll go 3-4 times and on a few occasions, we have even hosted small ensemble concerts in our home. Unfortunately, every time I turn on my system for a listen after having experienced one of these events, there is a bit of a letdown but such are the challenges of trying to recreate an unamplified acoustical performance because it's a fool's errand. For some time now, noise and harshness have not been the problem. What high power CPUs bring is a greater sense of being there. What lacks from the low power solutions that I have tried is scale and immediacy. Of course, it can be argued that these qualities fall more on the room, the speakers, amplifier, preamplifier and DAC but if the qualities you're seeking aren't adequately portrayed by the source, whether it be a music server or turntable, then those qualities are also compromised once they reach the speakers.
As for exactly why high power CPUs sound better to my ears, any answers that I offer would be purely speculative but the answers already provided by Emile are probably the best that I have seen. Yes, high power CPUs have the potential to create more noise which is why their use is counterintuitive but they are not the only source of noise in a music server. RAM is also a potential large source of noise and so when I asked Emile why he would use 48GB of RAM in the Extreme, this is what he said to me in a private e-mail and it really addresses both high power CPUs and RAM:
"Well RAM is a topic on its own, to start with, the 2 cpu’s are split into domains (NUMA / SNC), so you really have 2 x 6 dimms, 6 for each CPU, they are not shared. Music services have their own cpu/dimms and the OS has its own cpu/dims. So its sort of a core and endpoint into a single machine going beyond just core allocations for individual processes. These Ram modules are a custom order type, similar to the Apacer types popular in the Jplay forums, but taking it just a bit further. They do create less noise and draw less current then other offerings. If more dimms reduce performance, it typically means your power supply is negatively impacted by the increased current draw. As occupying more memory channels increases bandwidth and reduces wait states, you do get better individual process performance." "What you really want to do is reduce your hardware active processing times as much as possible. The net effect is much like a class A amplifier, you have a higher baseline power consumption, but power draw does not vary much, and this is very good for a more “natural/relaxed” sound. I hope this makes sense But you do need a power supply which is very comfortable supplying the load. You really want the least possible variation in load, and higher cpu power / bandwidth systems are better at that with very low load music playback processes."
This mirrors my own experience with regards to power supplies. As I experimented with endpoints with low power ARM processors and moved on to Celerons/Pentiums and then to embedded 4-core i7s within NUCs, a 6-core i7-8700T with a TDP of 35w, an i7-8700K with a TDP of 95w, and most recently a 12-core AMD Ryzen 3900X with a TDP of 105w, as I already posted, dynamics and staging improved but so did harshness. Like you, I have power supplies that have been custom made for me by Sean Jacobs. I have also tried supplies from HDPlex, Teddy Pardo, Sbooster, Uptone Audio, Kenneth Lau, and others but the one I prefer by a large margin are the dual regulated linear power supplies by Paul Hynes from Scotland. These supplies have a line rejection of about -150dB and using this type of supply to power my CPU, provided that I kept CPU temps <50 degrees C, harshness essentially disappeared to the extent that these high power CPUs now sound no harsher than the ARM-based endpoints that I have.
Now, to answer your question about correcting incoming noise before it reaches the server, yes, this is still important and like you, I am very OCD about this. I run multiple dedicated 20 and 30A lines to my listening room with 6awg specialized wire in my walls. I use high conductance outlets. I have tried AC regenerators including the PS Audio P20 (which I did not like) and have tried completely disassociating my system from mains with a Stromtank (which had no positive impact in my system). I have tried isolation transformers and balanced power supplies like the Equitech Son of Q. I have gone through a gamut of line conditioners including a Tchaik 6, Audience aR6-TSSOX, Shunyata Triton V3, Niagara 7000, and Synergistic PowerCell 12 UEF SE only to find that none of these came close to a Sound Application Reference TT-7 by Jim Weil. Power cords, don't get me started but basically I prefer the Pro Series cords from High Fidelity Cables. As for my network, my incoming cable internet feed is transformer isolated, my modem/router has had its clocks replaced by SOtM and is being powered by a Paul Hynes SR7. I am using stacked SOtM sNH-10G network switches (one is a dirty switch and the other is the clean switch) that until recently, were being clocked by a Mutec Ref10 master clock and these switches are also being powered by supplies from Paul Hynes. I have incorporated the Startech SFP modules that Emile recommends and so my 2 switches are presently connected by fiber. As for Ethernet cabling, you prefer the ones from Synergistic Research, my preference is for the SOtM dCBL-CAT7 but this will all be converted to fiber once my Extreme arrives. Despite all these measures, do high power CPUs still matter. Yes, empathically, yes.
