It's a question that gets asked and discussed frequently. USB is a lot faster than DIN, right? So why wouldn't USB be as good, if not better, for transporting MIDI Clock messages as MIDI DIN is?
Imagine there's a 1-way tunnel, and the cars are all traveling through that tunnel at 35 mph. And there are just enough cars wanting to use that tunnel so that they can all merge without slowing down or waiting. There's a constant rate of speed and, as long as not too many cars want to use the tunnel at once, or there's not that 1 car behaving badly, it's steady and predictable.
That's MIDI over DIN.
Now take the tunnel, make it 2-way and add gates at both the entrances and exits to control the flow of traffic. Put people on those gates that must radio each other to decide when to open the gates and let the cars through. Increase the speed the cars can travel to 1000 mph. Once the cars are allowed to go, they will get to the other end extremely fast.
That's MIDI over USB.
Overall, it's possible to get many, many more cars through the USB tunnel in an hour than the DIN tunnel, so by that measurement USB is far superior.
But, these analogies are missing one important thing: MIDI Clock is used for real-time sync of musical devices, requiring a steady stream of clock pulses, evenly spaced. Despite it's lack of speed, the slower, 1-way tunnel is actually better at this.
To make the faster cars passing through the USB tunnel appear to be coming through at a slower steady speed, they can be buffered as they come out of the tunnel. Basically, the cars arriving at the end of the tunnel wait a little longer, allowing enough time to be sure there's always a steady stream. But this is at the cost of real-time passage. It's adding what's known as latency. In most cases, latency goes entirely unnoticed. There's no perception you have that anything was delayed.
But what if your music equipment is trying to perform a time-sensitive task, like syncing devices together while musicians are performing? What if that buffer is causing clock pulses sent by one device to not arrive in real-time to the next device? Or even worse, when they do arrive, they are no longer as equally spaced as they were when they left (also known as "jitter")?
The answer to that question is the reason why USB can be a detriment for real-time musical requirements.
It's not all doom and gloom for USB though, there are situations when USB can manage fine. Admittedly, even though it's not as steady of a clock signal, there are situations where you may not notice it. In those situations, it's hard to beat the convenience that class-compliant MIDI USB has over MIDI DIN, simply because USB is ubiquitous and does not require the installation of more drivers.
But when it does fall apart, it really falls apart.
It's important to be aware of these things, and understand why USB tends to be mentioned as the least favored clock connection for MIDI.
A device such as CLOCKstep:MULTI provides connections for MIDI DIN, MIDI TRS, and MIDI USB, as well as an Audio Sync and +5 Volt Sync connections, not because we are promoting USB as an equal performer, but because we understand the convenience that comes with having choices. When asked, "Is USB good for MIDI Clock?" the answer you are going to hear from us is, "Not as good as the other choices, so use them instead if that's an option for you.".
P.S.: All that I've said here relates to MIDI 1.x, which has existed long before USB was a thing. If you are watching the developments of MIDI 2.0, you may be aware that there are new additions that are meant to help mitigate "jitter", but that's a conversation for another day.