As I understand it, the focus up 'til now has been correctness and compatibility rather than performance / efficiency. It seems a bit early to jump on them for that (though the speed hit you mention is surprising). That said, I'd expect things to pick up speed once the next-generation compiler Laurent mentions in the blog post is ready (and when the RubySpec needle moves closer to 100).

Regardless, the team has made a ton of progress since the last release. Here's looking forward to the next.