Re: Apple Studio Display (2026, Thunderbolt 5) — D P tunnel torn down every few minutes on Intel TB4 host

Hi Mika,

> This is not expected. What typically happens is that there is another
> hotplug and the DP tunnels get re-established.

You are right and my observation was wrong — I had queried tbtunnels after
the re-establishment, not during the outage. It goes exactly as you describe,
about a second later:

  16:47:48  acking hot unplug event on 1:11
  16:47:48  1:11: DP OUT resource unavailable: adapter unplug
  16:47:48  0:5 <-> 1:11 (DP): deactivating
  16:47:48  acking hot unplug event on 1:12
  16:47:48  1:12: DP OUT resource unavailable: adapter unplug
  16:47:48  0:6 <-> 1:12 (DP): deactivating
  16:47:48  looking for DP IN <-> DP OUT pairs:
  16:47:48  0:5: no suitable DP OUT adapter available, not tunneling
  16:47:48  0:6: no suitable DP OUT adapter available, not tunneling
  16:47:49  1:11: DP OUT resource available after hotplug
  16:47:49  available bandwidth for new DP tunnel 34650/34650 Mb/s
  16:47:49  0:5 <-> 1:11 (DP): activating
  16:47:49  0:5 <-> 1:11 (DP): DP IN maximum supported bandwidth 8100
Mb/s x4 = 25920 Mb/s
  16:47:49  activating Video path from 0:5 to 1:11
  16:47:49  Video path activation complete
  16:47:50  1:12: DP OUT resource available after hotplug

So: unplug on both adapters, teardown, then a fresh hotplug on both and full
re-establishment roughly one second later. Nothing stale. Sorry for the noise.


Cable and power supply
----------------------

> Also, you connect the monitor with a real TB cable, right?
> The second idea that comes to mind is power supply

Both original Apple parts, and I can answer this better than by description,
because of how the desk is arranged: the NUC sits physically on top of a Mac
mini M2 Pro. Same room, same mains outlet, same monitor, same cable.

  cable plugged into the Mac mini underneath  ->  stable
  same cable plugged into the NUC above       ->  teardowns within minutes

Nothing moves except which host the plug goes into. That rules out the cable,
the monitor's power supply, and anything environmental. Note the Mac mini M2
Pro is Thunderbolt 4 as well, not 5, so this is not a TB5-sink-on-TB4-host
problem either.

The cable was additionally swapped for the 2022 unit's Thunderbolt cable
earlier; link still came up at generation 4 and the failure was unchanged.


> Based on your dump it also happens to both DP OUT adapters (11, 12) at the
> same time.

Confirmed, always both, in the same millisecond bracket. The tunnelled USB3
and PCIe stay up throughout — only DP goes down. The monitor's USB hub,
webcam and speakers never disconnect during an event.


Usage pattern
-------------

> You don't see any "usage" pattern there when this happens?

Partly. Scrolling bright web pages full screen raises the rate noticeably and
is my usual way to provoke it. But it also happens with the machine idle and
nobody in the room, at a lower rate. So load correlates but is not required.

I am sorry to say I cannot run any more tests: the monitor goes back
today, so this
hardware is gone as of this afternoon and no further tests are possible on it.

What I do still have is everything captured while it was here — the traces
already sent, matching dmesg with drm.debug=0x104 plus thunderbolt dyndbg,
and full journals across several sessions. If any of that is worth a closer
look, or if you want a specific window extracted, just ask and I will dig it
out.


Bernd