Sure!
I started out many years ago wondering how a ripped CD would compare to my modified CD player (TDA 1547 gold crown without OS, TX2575 I/V resistor, SRPP E80C tube stage etc), so I started using a laptop and a Teradak based on the 1547 that I (again) modded the heck out, and never played another CD.
Then I wondered if a dedicated computer with a decent power supply would improve sound, so I got an intel NUC (dual core celeron) and a lineair 19V PSU...that kicked off all of this as tinkering with yet another power supply for the SSD hard drive was another huge improvement.
A 'few' experiments and years later I started the recent development path....with an AMD 6 core CPU on a budget MB using a standard computer SMPS, and was impressed...but got curious about the why (story of my life). So I got a better and more powerful 1200W SMPS, a CPU with 8 cores, a Pink Faun I2S output PCIe card on a linear PSU, and implemented passive cooling (funny enough is vibration a big lust killer in digital), which was the setup at BAM!
That led to some iterations with ever increasing core count, and the notion that more is better, different types of RAM sound different, more power is better...so I built a linear unregulated PSU around the time Taiko Audio released a DIY ATX module based on GaN mosfets.
(Taiko develops digital music servers, and I tend to think they are the cutting edge for music servers.)
I bought their ATX, started playing with active recifiers, got custom made chokes in Hi-b cut core by Martin Mug as there is a clear relation between instant power delivery and quality of sound, tinkered with 8 core , 12 core and then 16 core AMD CPU's and numerous wires and bypass caps.
Ever increasing detail, power, presentation, basically sound got better and better and more natural and dynamic...
in parallel there was a group of similar fools doing similar things in computer audio fora, and they veered off into a dual Xeon build, sort of a poor mans Taiko Extreme..
At some point a 16 core CPU was about the max I could get, the next level was server CPUs like the new AMD threadripper, and Intel Xeon....and knowing Taiko did audition about any relevant MB (and had one custom built) and CPU there is when they created the Extreme I took the jump and started the current build based on belief they know what they are doing planning this to be the über build
Long story short, more computing power means better detail retrieval (so it IS there in that digital, even when streaming), vastly better dynamics and weight, top end becomes sweet, leading edge trainling edge AND texture of sound in the middle.
Something similar goes for power supply....it needs to be capable of delivering power in femtoseconds...a CPU is fed by 12V to the MB where it is turned into approx 1Volt that goes direct to the CPU pins..the faster that 1V can deliver current the better stuff sounds.. so that is likely why MB with more power phases do better for audio.
SSD is another thing...there is a difference in what SSD data is stored on, with Intel Optane and pseudo SLC being best IMHO, heck even what the operating system is stored on makes a difference in sound, what OS matters for sound too. Same with RAM.
Current setup (work in progress):
Audio router and switch on 2 linear PSUs with active rectifying, Lundahl chokes, loads of mundorf 4 pin AG caps and Sean Jacobs DC3.
Dual Xeon 4210 on Asus Sage C621 and 12x4 Gb ATP RDIMM running Daphile (Arch Linux distro with LMS integrated)
Solarflare X2522 plus (ultra low latency) fiberoptic network input, finisar 1475 BTL SFPs using corning glass fiber with grade A connectors and
Pink Faun I2S bridge with Ultra OCXO
Active rectifier using GaN, Mug Hib cut core chokes, Mundorf AC and HC caps, numerous fancy bypass caps
All in a pertinax box
playing into my Metrum Adagio DAC 3
basically the game is to have the CPU's do as little as possible, a CPU is as apparently inherently as lazy as we humans using them....CPU 1 handles a couple of tasks, CPU 2 some others....just pushing bytes from the stream or the NVME SSD with music files takes nothing (O and do play from memory and store as redbook WAV, not FLAC, no high res unless that is a native high res recording etc),...no upsampling, no filtering, no room correction....CPU load as low as possible, any functions that are not required shut down in BIOS, it is all clearly audible.
Bottom line is, a rapberry Pi can do the same thing, but this sounds so much better.
