I should have added, yes, the Cray-1 and later were vector processors (in favour until the early 90's) for numerical work, even then Seymour was reluctantly abandoning this approach (likely starting with his earlier STAR work at CDC), he was vector obsessed.
I think by the 80's IBM managed to microcode the S/370 instruction set on a pair of 68000's cumulating in the PC/370 and AT/370... think it went further than that as well. I know by 2000 there was a desktop S/370 capable of running VM, MVS etc.