Why does USB feature so much in discussions about DACs when the newer HDMI seems better?

HDMI isn't really the correct terminology for audio only use. Yes, the connectors are HDMI, the cable itself needs only be I2S, this means only 8 pins actually wired, the rest are superfluous for audio only. Plenty of nice audio only I2s cables out there today, Tubulus and Audiolund are a couple manufacturers.

I2S has a couple advantages over usb in that data and clock runs are segregated, usb has to extract data from clock run. The other advantage is I2S is native signal path within dacs, usb inputs converted to I2S. Having said that manufacturers have gone to great lengths optimizing usb at both the streamer and dac, so usb can be very nice interface these days. Theoretically, I2S imputs on dacs shouldn't require as much work as usb, still clock and power supply to that clock should be optimized. I2S is virtually unknown output on streamers, and if included not likely to be optimized. So now we come to the greatest liability of I2S, that being not designed to be sent over cable, clock should be located in closest proximity to I2S circuit within dac. For this reason best to limit I2S cable length to .5M or shorter, using such a short cable could be problematic for many due to equipment positioning issues.

And then we have the question of optimization in the implementation of usb and I2S, such that a well implemented usb would likely better less than optimized I2S. Top of pyramid would be quality implemented OXCO clocking on either of these interfaces, this mean quality clock and power supply to that clock. Next would be TXCO, then Femto, this hiearchy not set in stone, therefore, quality implemented TXCO or even Femto could better less than optimal OXCO.

Bottom line, so many variables, individual setups could favor usb or I2S. In my case I've come to the determination I2S superior to usb since I have both well implemented usb and I2S to choose from, I2S beats out usb. I also maintain I2S inherently superior to usb based on reasoning explained above, this as long as I2S cable length .5M or less.

'Experts' have a variety of opinions in this matter, some refuse to offer I2S dac inputs based on a particular mindset, others offer them based on another set of reasoning. No one I'm aware of offers optimal I2S input in dac, OXCO clocks with high quality power supplies simply take up too much real estate to offer in dacs. Optimal I2S rarely if ever offered on streamers either, usb seems to be their de facto output, sometimes network. This is where ddc come in, they take what is hopefully optimized usb out of streamer, convert to quality I2S that is synced or slaved to dac, this means one is using clock and power supply from ddc vs dac internal clock and power supply. There are a small number of custom streamers such as mine able to directly output well implemented I2S to dac, this means bypassing usb altogether, this may be superior solution.

Finally we have the proprietary interfaces like the ones in Auralic, Taiko,etc. based on reviews these surpass both usb and I2S.

Here I am quoting from a respected DAC designer:

OK...see if this makes any sense to you: if clocking gets corrupted with a single channel traveling on one wire then how would it make any sense to attempt to coordinate three separate clocks on three separate wires?

It makes no sense.

If I2S was actually better they would be using it in recording studios and they most certainly do not.

If I2S was actually better then nearly every company in the audiophile industry would be promoting it and they most certainly do not.

There are a small group of Chi-Fi manufactures who started promoting I2S and the audio-fools bought into it hook-line-and-sinker.

If I2S sounds better in a specific DAC it is only because the other digital inputs on that DAC are lacking, not because I2S is inherently better.

The above excerpt is from pg 7 of this excellent thread.

Really, it is puzzling to see audiophiles bicker over various types of digital connections that are either deprecated (spdif/toslink/coax), never meant to operate over cable (I2S), or that can be made to work well by jumping through a thousand hoops (USB).

By the way, why are audiophiles still stuck in USB 2.x? The 3.x revision has nine wires (vs. 4), enough to carry more clocks than a train station if that's what you think will help sound quality. 

Meanwhile, AoIP - Audio over IP - is a thing, and has been for years. Runs flawlessly over standard RJ-45 or better yet, SFP. Dante and Ravenna are well known AoIP implementations popular in pro audio. But to my limited knowledge there has been a grand total of one Ravenna-enabled audiophile DAC. Why?

@devinplombier - very, very few people understand how digital interfaces work. It is easier to understand spdif/toslink with all its jitter and whatnot, but explain USB or Ethernet... good luck. I often feel people believe there is an actual "stream" of bits between Tidal server and their device...

@sns HDMI natively supports pure audio - have a look at my earlier post where I list what it can do.  It is a very impressive list, certainly way beyond I2S.

Now, HDMI connectors are attractive for other uses.  I have a camera system in my motorhome which supports four analogue TV cameras feeding a single display. Dometic, who makes this system, choose to use HDMI connectors to carry the analogue TV signals.  Despite there being a standard HDMI connector for in-vehicle use, Dometic use the totally unsuitable consumer connector!  Hopeless.

I had never heard of I2S when I started this thread, but I can understand how HDMI with its 19 connection pins would provide an easily available way to transmit four signals.  But it certainly is not native HDMI.  There are 10s of billions of HDMI connectors in the world today.