Performance differences between RID Lookup vs Key Lookup?

Are there any performance differences between when a non-clustered index uses the clustered index's key to locate the row vs when that table doesn't have a clustered index and the non-clustered index locates the row via the RID? Does different levels of fragmentation impact this performance comparison as well? (E.g. in both scenarios the tables are 0% fragmented, vs 50%, vs 100%.)