I'm not sure about any current Perl6 vs other language benchmarks; older benchmarks aren't terribly useful because within the past year MoarVM both became the dominant P6 VM and also received substantial performance optimizations. My understanding is that this is one of the big goals of this year.
We try to :-). Specifically, there are designs underway to implement e.g. hyper operators - those things that apply an operator to an array of values, irrespective of order - by SIMD instructions on the hardware level. If you're interested you're very welcome to come and check it out.
[edit] - there is also some progress in speeding up list operations, because as it turns out making all lists lazy is not a really good way to do fast list computations. This is known as the 'Great List Refactor' (GLR) in perl6 circles, and spoken of with awe :-)
We don't really use a mailing list, but we do have an IRC channel on freenode: #moarvm. Although just as often MoarVM-specific issues are discussed on #perl6
In his talk, Larry specifically said something like "Please understand that we may or may not be comparable to Perl5 in some benchmarks, and that speed is something we're working on. We'll get there, and it's good enough already for many tasks."
Another one of Johnathon's excellent presentations goes into depth about the kinds of optimizations being made in MoarVM: http://jnthn.net/papers/2014-yapceu-performance.pdf