See also the full set of language benchmarks: http://attractivechaos.github.com/plb/

Interesting to see V8 performing so well on speed in all benchmarks. JaegarMonkey seems to use 4x less memory in the matrix multiplication benchmark though!