Just as a thought experiment: what sort of attack could Intel make if you limited your instruction set to the bare instructions (i.e. no special crypto assembly

Just as a thought experiment: what sort of attack could Intel make if you limited your instruction set to the bare instructions (i.e. no special crypto assembly code)? Wouldn't it be unfeasible to attempt to detect crypto operations using normal assembly functions without even knowing whether crypto is taking place or not?