Use low gain unless you really need more power. The Atom measures a lot better in low gain than it does in high gain. There is a little study on ASR about gains which compares the Magni 3 and the Atom in their different gains. To lazy to link it now but you will find it anyway
I'd be surprised if it can drive
@iazriel's DT 770s at low gain (even with dedicated DAC, as others suggested in the thread) considering that my DX3 has no headroom at high gain for the similarly hard-to-drive DT 990s; not sure how much power the Atom outputs at low, but at high it's about double what the DX3 does at high, so yeah.
I think I found the post you're referring to
here; there's a comment from solderdude there that explains things (search on page 1). It seems that for hard to drive headphones it's unlikely you're ever going to get any audible distortion on high gain, if I understand correctly that is. So I would maintain my advice that
@iazriel will be best served in high gain since they've got insensitive headphones like mine, even if they buy a DAC (although I'm sure that will improve the situation a bit).
This is a question of priorities and your system. If you're gaming, I would particularly avoid USB 2.0 because of latency, which may be felt as additional 'input lag' and possibly audio sync issues and stutters. USB 3.0 is less of a problem (I think, but not 100% certain) and so the S/PDIF may be preferable. You have to make a judgement call on how powerful your system is vs the number of devices competing for time on the USB. If you're not gaming, use whichever is the most convenient.
Eh, I don't think this could ever be a problem even for computers that are like 20 years old, nevermind modern ones. I've never seen any evidence to back up the various talk about USB 2.0's latency/polling problems in the real world (neither with audio, or with keyboards which is a claim you often hear from people who prefer PS/2 keyboards that use actual interrupts). I mean from a hardware perspective, USB 2.0 is absolutely an inferior bus to 3.0, but even so, I've never seen data to back up that it actually makes a difference in the real world. I think people underestimate just how overkill a modern CPU is for something as simple as USB. Also, if I'm not mistaken, it's the host controller that does the polling, not the CPU.
Lastly, 99% of DACs are USB 2.0, even for high-end professional gear. I've even seen at least one DAC that had USB 3.0 ports that was actually USB 2.0 internally and had claims on the vendor's website about how they have the 3.0 ports just for the sake of supporting USB 3.0 cables which have higher shielding requirements. I can't imagine you getting any benefit from plugging a USB 2.0 device into a USB 3.0 port; you'd be using XHCI instead of EHCI, but again, I haven't seen any info out there that suggests this has benefits.