The problem is, I'm not sure how it'd help, when you have to do emulation. It often requires explicit hardware support (or some system level support) to break ABI to not manually do mappings and stuff, to make emulation as little overhead as possible. Make sure you caught that, "as little overhead as possible", I am not good at hardware whatsoever but to me there's no reason that an alternative chip would be any better than the original chip being emulated, when you have the same ISA being ran....... UNLESS there is some recompilation. This is why I'm a fan of JITs and their little VMs, you can do your own JIT compiler for VMs and implement your own optimizations, the same way you'd, say, implement fast-paths in a JS JIT compiler.