I don’t have a clue really. Nvidia GPUs also have the new tensor cores that handle some workloads.

An easier comparison is probably to something like the amd ryzen processors where the integrated gpu has 7 or 8 vega cores. Each of these is likely at least in the same order of magnitude of performance as the m-series gpu cores.

CUDA cores are very different from other GPU cores. They will likely do at most 2 and normally 1 operation per clock cycle. These GPU cores will be doing dozens of operations per clock cycle.

You are saying a bowl of rice is more food than a massive steak since you get hundreds of rice in the bowl compared to one steak.