Well - IT IS RUNNING
tl;dr: It runs similar to a Pi 4B for CPU tasks but GUI is worse than VNC over a poor WiFi link
I made a slight change to the VM - I forced it to have 2 CPUs but left it as a pc-q35-7.2 (whatever that means).
No real idea why it was looping but I powered off the VM and this time when starting it up I did not use 'pi' as a user and everything went through.
Response is not great in a window (still to test full screen etc) but CPU tasks are not bad (compared to a Pi 4B) using the basic sysbench 1.0.20 test:
Pi Desktop for Mac / PC under UTM (2 'processors / 4GB):
- CPU Events per second 74.98
- Total time 10.0099s
- Total events 751
- Latency 11.20 / 13.30 / 78.90 / 25.25 / 9988.73
- Threads fairness 751 / 9.9887
A Pi 4B (8GB 64bit Buster):
- CPU Events per second 118.76
- Total time 10.0056s
- Total events 1189
- Latency 8.33 / 8.41 / 25.24 / 8.43 / 10001.67
- Threads fairness 1189 / 10.0017
A Debian ARM64 Buster VM under Parallels (2 'processors' / 4GB / with Parallels tools installed):
- CPU Events per second 11011.56
- Total time 10.0001s
- Total events 110122
- Latency 0.09 / 0.09 / 0.90 / 0.09 / 9988.48
- Threads fairness 110122 / 9.885
I will see about setting up a native ARM Debian under UTM at some point soon and add the scores here.
edit: Set up a Debian ARM64 under UTM (2 'processors' / 4GB / Not using the Apple Virtualization backend):
- CPU Events per second 10980.64
- Total time 10.0001s
- Total events 109813
- Latency 0.09 / 0.09 / 5.15 / 0.10 / 9989.15
- Threads fairness 109813 / 9.9892
Close enough performance wise for me to continue testing UTM before my licence runs out on Parallels...
Interestingly, UTM again tried to go through the install till I powered off the machine and started it again. I think in this case it is not ejecting the CD but that does not explain the 'first boot' set up on the Pi image.