Still, running 2 to 4 5090 will beat anything Apple has to offer for both inference and training.
That won’t work for the home hobbyist 2.4KW of GPU alone plus a 350W threadripper pro with enough PCIe lanes to feed them. You’re looking at close to twice the average US household electricity circuit’s capacity just to run the machine under load.
A cluster of 4 Apple’s M3 ultra Mac studios by comparisons will consume near 1100W under load.
I would say 1-2 RTX 6000 Pro maxQ are more practical.