yamamoto2002

I think modern PCIe sound cards are marketed for gaming, watching movie or professional realtime music production, not audiophile music listening
 
Benefits of PCIe sound card over USB audio devices
Lower latency : It contributes competitive gaming, lower sound latency means more wins. Sound monitoring feedback for musician requires 10ms or smaller latency. I read somewhere, on some special hardware/software combination, it is possible to achieve ～1ms latency with PCIe cards.  Stability on high CPU usage : Game or Live/realtime sound production such as Ableton or Pro Tools consume CPU power and PCIe audio is less prone to happen buffer underflow sound glitch/stuttering on high CPU usage scenario, thanks to more down to metal implementation. Multichannel output for movie watching: It is related to #1, lower latency means less prone to exhibit lip sync problems. Many PCIe sound cards have 7.1ch analog out. But more serious movie fans use HDMI and dolby atmos AV amp rather than PCIe sound card  
Benefits of USB audio over PCIe sound cards
Standardized device driver: Windows provides USB Audio Class device driver and it provides plug and play experience, just plugging USB audio to start using it, No need to install IHV provided proprietary device driver which is sometimes unstable and also sometimes it includes bloatware. Easier to countermeasure EMI noise problems: PCIe cards are usually installed onto inside of PC box and there is a variety of electromagnetic noise source there. USB Audio devices is connected external port of shielded PC case and there are many products to galvanize PC noise.

IMO there is no or little benefit on home network but it is okay to do so.
 
Link aggregation provides redundancy and throughput.
 
Redundancy: when one cable of LAG is failed, network will continue to work, as you said.
 
Throughput: (Scenario 1) when two client computers simultaneously read/write to single file server, link aggregation may improve total throughput of the file server (only when you are lucky). (Scenario 2) but, when the both two computers are connected to hub with link aggregation, single file transfer uses only one link and transfer speed between this two computer is not increased.
 

 

 
 
(About Scenario 3 slide) please be careful: if two cables are connected between two hubs without proper switch lag configuration, loop is formed and it may cause network breakdown

Windows 10 privides several different sample rate converters to resample PCM signal for apps, with different conversion quality. Also apps may have own sample rate converter.
 
I don't know Amazon HD uses which API to play sound, but SRC can be bypassed when source sample rate and shared mode sample rate is matched.
 
Shared mode's sound altering filters can be disabled on Enhancement tab of Sound Control Panel (Fig. 2).
 
When SRC is bypassed, remaining problem of altering sound of Windows shared mode is Limiter APO, which reduces gain temporary when incoming PCM is too hot. It starts to work when incoming sound magnitude exceeds 98.5% of full swing.

Windows provides these sample rate converters:
MME Resampler: Used when App play sound using old MME API. I don't check its conversion quality of Windows 10 versions. Direct Sound Resampler: Short-delay (not linear-phase) resampler. Conversion quality is low. Media Foundation Resampler: Latest converter of Windows. Linear-phase. Conversion quality can be controlled by app. App's own resampler: Music player apps may have their own resamplers for their needs.
 
I believe all SRCs of Windows 10 provides use CPU.
Most app's own SRC use CPU, small number of apps use GPU to perform SRC.
Also there are audio hardwares such as Lynx AES16e-SRC that contain hardware sample rate converter chip.
 

Fig.1
 

Fig.2 Enhancement tab of Sound Control Panel
 

I use Adobe Audition de-clippper function when crackle noise is heard from 1990s or earlier classical recordings. It seems iZotope sells more comprehensive audio repair tool

Ethernet switch should speak AVB protocol otherwise AVB system does not work as designed. AVB system consists of AVB talker (sender), AVB bridge (switch) and AVB listener (receiver). AVB participants are synchronized with one master clock without word clock cable using IEEE 802.1AS time sync protocol. IEEE 802.1 Qat protocol provides the mechanism to reserve some bandwidth for audio stream, something similar to isochronous mechanism of USB. IEEE 802.1Qav guarantees worst case packet forward delay.
 
It seems some models of Cisco Catalyst 3850 and Catalyst 3650 also support AVB protocol.