Has anyone moved from a dedicated streamer and/or server to a computer-based setup with separated components? My current setup uses a streamer/server all-in-one and I’m just wondering if anyone has made the transition from a similar situation to multi-box solution?
I’m aware that the biggest benefits to a 1-device-based system is convenience and footprint. Some could also argue the software for dedicated all-in-one components is a plus as well. But has anyone trashed their single device to venture out into fiber cables, interfaces, separate power supplies, etc? If so, why?
I have 2.
A PC I built (W10) that is only used in the system, and a Pi 4 that I built to piddle with i2s using Iancanada’s FifoPi to reclock and HDMIPi for the i2s/LVDS output.
I run JRiver on both. Both appear as renderers (via JRiver) in BubbleUpnp. Both (along with Bubble) can also play to the Bridge II as well.
I’ve been piddling with the Douk U2 Pro, and just got a deal on a Matrix SPDIF 2 to piddle with as well and to compare i2s via USB with the Pi4, which outputs i2s directly through the hats.
I’ve never had a “dedicated streamer,” and most (not all) are either standard motherboards (usually NUCs or even Raspberry Pis internally (such as the Mano (Pi3), Mercury Streamer (Raspberry Pi Compute Module 4 on a proprietary output board), etc.).
I debated about the PinkFaun pcie card, but decided it wasn’t worth the price, I actually tried it years ago (not current version) and had to return it because it didn’t work with the PWD at the time (I’m sure that’s fixed now), it’s not as fun, and I believe it can’t do DSD (even though I really don’t have any DSD files I listen to - just play with).
Before getting the server, I tried out JRiver through a Macbook and went straight USB into my DAC. The mids were pretty harsh, and the ease of storing everything in a single neat place tempted me. I’m pleased with having a dedicated server, but I always remain inquisitive. After visiting this thread several times, hearing some of the success stories and learning more about how people are successfully doing what I’d gotten wrong earlier, I thought of dipping back into those waters.
When using JRiver, for ‘best sound,’ you need to make sure you set it up for “bit-perfect” transmission. There are so many setting in JRiver from DSP options to buffering options, it’s easy to muck up the data stream. In certain skins such as the Black on Black and Noire, you can confirm bit-perfect playback by making sure the audio path icon is blue while playing, indicating JRiver is not altering the data from the original file in any way.
If not JRiver, there is a plethora of options - Roon, etc., along with free options as well - Foobar, MediaMonkey, etc.
I have yet to go in the server direction.
For what it’s worth, my iMac/Wi-Fi/Router/Ethernet-Optical/PSA Bridge II/PSA DSD DAC system is an excellent performer.
JRMC is my current, preferred rendering software (second to Roon), and the system is very, very stable.
I am sure it can be bettered, but all systems can be improved upon.
[PS, my bias is that USB is an inherently inferior means of taking the 1s and 0s out of a computer when the digits represent music and have to be “streamed” to a DAC. Since I have never used USB for that purpose, it MAY be why I have not experienced the harshness you recall. 
]