Mmmm, not really. I have both a4x 3090 box and a Mac m1 with 64 gb. I find that the Mac performs about the same as a 2x 3090. That’s nothing stellar, but you can run 70b models at decent quants with moderate context windows. Definitely useful for a lot of stuff.
>quants
>moderate context windows
Really had to modify the problem to make it seem equal? Not that quants are that bad, but the context windows thing is the difference between useful and not useful.