No, you need utf8 to have perl understand literal utf8 in your source, even in 5.8. (I think you can use a BOM instead, but I don't know for sure.)

The character is U+00E9. Perl may choose to store it internally as C3 A9 (flagged as utf8) or just E9 (not flagged as utf8, and assuming appropriate locale). Can you clarify what you mean by "This is not what I want."? Perl certainly didn't output that "65533" that you show. May I suggest you change your test to show you what is in various variables using Data::Dumper (preferably with $Data::Dumper::Useqq=1)?