MBA 2019 GPU/Logiboard hanging even after 4th! logic board! (IOAccelFenceMachine::fence_timeout)

lxmeta

macrumors newbie
Original poster
Dec 6, 2018
26
15
Austria
I have had troubles with a 2019 MBA out of the box.
Initially I was barely able to connect TB3 devices without issues (freeze, reboot...)
Now I am on my 4th logic board, and the problem has slightly eased (less immediate shutdowns) but still unusable with an external TB3 monitor, even when using an eGPU.
The message I keep getting in the log:
IOAccelFenceMachine::fence_timeout(IOTimerEventSource *): IntelAccelerator prodding blockFenceInterrupt
or for the eGPU:
IOAccelFenceMachine::fence_timeout(IOTimerEventSource *): AMDRadeonAccelerator prodding blockFenceInterrupt

Now on the 4th logic board and Catalina 10.15.3 the issue is slightly less bad, but still catastrophic.
  • Sidecar with a cable is now working for the 1st time! Beforehand it stopped immediately with an error after connecting for a few seconds. Note: This is USB3 directly connected to the TB3 port
  • The "apple recommended" Gigabite eGPU is now being detected, and the monitor shows the screen! (Not the 1st board!)
  • The sporadic crashes are gone. I had panic reboots with and without backtrace on the first 3 boards

Problems are still simple to reproduce:
  • If you have a TB3 device connected, that has a USB(3) hub integrated, and you tunnel something equivalent of a full hd camera to the GPU to work on it (video decoding), system becomes unresponsive. This is simply a TB3 Monitor with a built in fullhd camera. This worked without any issue for 3 years on a MacBook Pro 13"!
  • When the error message starts popping up in the console, the audio and bluetooth connections are dropping. Not only the AirPods connected through BT are dropping, but also the cable based apple headphones are "getting lost"
  • The google hangout video chat now starts switching between audio sources like crazy, as they are becoming unavailable/available again. This is also true in the system settings (if you still can get there!).
  • I had cases where I am getting the error message "IOAccelEventMachine2::hardwareErrorEvent() : GPURestartEventDequeued stampId=9 type=3" following a message that the Intel GPU went down...
My verdict is simple: This specific iteration of the MacBook Air is faulty by design, while as usual Apple is not responding or admitting anything, but trying to put bandages on a design error.
Adding an eGPU to fix the "GPU is too weak" issue does not help, as the problem apparently is also with the bandwidth between the CPU power and the IO bus to all the devices connected ...

Anyone having the same issue? I can barely believe that it's only me, unless everybody else accepted that the official specs of the MBA 2019 are just fake news...