There is a buffer on USB and it is an asynchronous transfer technology, which is what creates the latency. Direct path to CPU etc. doesn't mean much as the USB latency itself is minimal, it is the buffering on both sides that are the main bottleneck. If you have 1024 samples of buffer then you have aprox. 20ms of lag just on the transmitting end. Then add long filters in the Chord DAC etc. and you get quite a bit of latency.
In a low latency game on a low latency system the delay is typically 20-30ms from peripheral to on-screen action. Low latency monitors are typically 3-7ms of latency while "normal" monitors will typically have 15-30ms of latency. The delay on the mojo makes the sound noticeably out of sync so I wouldn't use it over USB for any kind of fast paced gaming. It is fine for slower more casual gaming where timing isn't that important.
You can still get ground plane interference as there is still a reference ground between the units, your ground is just floating. Only way to have proper galvanic isolation is to use something like an Intona or similar device or to have galvanic isolation in the USB port of the DAC (not feasible on a unit that must work from phones due to powerdraw).
I can only control what I can control. I cannot change nor control the overall USB design. And with any design, there's likely compromises along the way. Direct path to the CPU makes a difference. Tell that to Nvidia or AMD that a direct path to the CPU is minimal. Try running your Graphics Card on a non-PCIe connection and let me know if the difference is minimal. Data is data. Shortest path optimal. There are Audiophile USB devices that cater to having a direct path (PCIe) to the CPU.
For example here's a $1000 USB device. Cheap, generic clocks so I would never use but just a quick example:
I get your trying to point out in theory that this may be an issue, but in practice it's not. I'm glad you raised these points though.
As for Galvanic Isolation. I started thinking about this after my Soundstage improvements. I'm actually running Bulletproof Galvanic Isolation. 100% free of any Ground Plane Interference because it's decoupled by optical. 100% free of any power issues because I'm running completely Off Mains so decoupled and I'm running the best unregulated native 5V modern power supply possible with almost zero impedance. My 5V power supply destroys a car battery even if a car battery was native 5V. That's what I'm using to power my USB optical cable.
Yes a car battery is the best possible PSU - huge dynamic current, ultra low impedance and no noise. But the real benefit is batteries, and isolation from the mains, so a car battery is not needed; just use a USB power bank.
https://www.head-fi.org/threads/watts-up.800264/post-15600310
For Chord Table Top Galvanic Isolation, it's not bulletproof. There's potential leakage. That's why some whom had high hopes because their Table Top DAC has "Galvanic Isolation" moved on from USB to Toslink because it's not bulletproof. Fortunately, I can stick with USB because I overcame this leakage issue.
Intona, etc. == gimmick. I'm about having a fundamentally "Off Mains" sound chain, not Marketing chains. But it takes years of Rob Watts teachings and Chord experimentation to keep things simple and fundamentally correct, so I understand it's not easy for most bombarded by Marketing propaganda and mixing in something that may work for other brands. We all go through different Chord stages with lessons to learn.
Chord users have their own Chord ecosystem so they try to do what's optimal and fundamentally sound for Chord products. Third-party products usually don't work out well especially those that are not optimal for Chord.
There's nothing out there that is close to my bulletproof galvanic isolation USB chain. I realised this only after my Soundstage improvements last month. It's not something I set out to accomplish or cared much about when I started this project. It happened naturally. Heck, I just started with I possibly have "Galvanic Isolation" theory a few days ago thanks to posters here.
But I do appreciate your response as I value other perspectives.
EDIT: Ir-regardless of USB Buffers, even read and writes dramatically outperform on hard drives, SSDs, NVMEs, etc via PCIe (Direct connection). In practice, the USB buffer latency is a non-issue whether regular USB data or USB "audio" data. The performance with a direct connection to the CPU is night and day. Graphics cards would come to a halt without a direct connection to the CPU. For users of these high performing direct connection to CPU that experience tremendous hard drives speeds, I don't think they worry too much about USB Buffers.