[bitL]

Yes, Nvidia still obliterates M1/M2 in Deep Learning. M1 is close to GTX1650 in real-world DL workloads though in theory based on TFlops it should be around GTX1070.