[j45]

The cost of maxing out a mini from Apple typically puts it at least double the cost in a computational power per watt model.

It might be feasible to but 2 or 3 usffs to cover one max mini at a fraction of the cost, or stick the mini to doing only certain tasks on 8 or 16gb.

Since the m1 addresses memory differently less ram should go further, but I’m not sure if those efficiencies extend to virtualized machines.

I’ll try to dig up an old spreadsheet and add the the m1’s and m2’s stack up to the above.

The arm docker image is a real deterrent. Forums have more and more workarounds and tweaks so hopefully they’ll become available as time goes on. It’s often not worth fighting with compiling.