Rubinius can run pure ruby code much faster than YARV because of things like a better GC and the ability to compile ruby code down to machine code (a JIT compiler).
Rubinius is much easier to work and understand than YARV because much of the core functionality is written in Ruby itself. This lends itself to being easier to hack on and add features to.
Rubinius homepage claims that it's fast but has no benchmark results that I could find, which makes it hard to take those claims seriously.
I like the way JavaScript engines are being compared using two or three standard benchmarks.
(to pre-empt "benchmarks are useless" comments: yes, benchmarks are flawed because they only measure what they measure, but a decent benchmark is better than no benchmark).
It's much faster at executing pure-ruby code. Since it has a much smaller portion implemented in C, total execution time may or may not be slower than YARV; it depends on the execution profile of the task at hand. Code that spends most of its time in things that YARV does in C will be slower on Rubinius.
It shows 1.9.2 to be a bit faster most of the time, and to have a smaller footprint. Still, Rubinius could be the better choice in some cases. The best thing is to run your own perf tests specific to your application, and rvm makes that head to head comparison easier to carry out.
Any word on 1.9 syntax support? Sorry, I havent been keeping up on Rubinius progress recently. I've switched all my projects to 1.9 syntax though so going back to 1.8 would be kind of a drag.
I switched to Rubinius earlier this year for day-to-day development. Good project and deserves to be supported. One issue for me however that is also the case for JRuby is that much more memory is usually used than with Ruby 1.9.2. If you need several Ruby processes running on a low end server to support a web app then memory footprint size is important.
All that said, I love the idea of Rubinius and it is an important project.
