In Java's early days, there was a technical reason for the difference. Do you know what a "virtual table" or "vtbl" is? Basically, a simple vtbl could be used for directly inherited methods, and so the JVM just had to look at two pointers to find the code for a method inherited from an abstract class -- that's relatively fast. But since any class can implement any number of interfaces, finding the code that implements an interface method would involve following a pointer to a table of interfaces implemented by a class, then searching through the table to find a vtbl for the implementation of that interface, and then finding the method pointer in the table. That's obviously a lot more work, so calling an interface method used to be measurably slower.
These days, JVMs are a lot smarter, and most of this sort of lookup is done during dynamic compilation, so the runtime differences are small or nonexistent.