The sunspider tests are also fairly dubious as a performance indicator because of their very short execution time, the Chrome team posted a modification which ran it 50 times to get a better idea of steady-state performance (http://blog.chromium.org/2011/05/updating-javascript-benchma...). Sunspider does act as a fairly good indicator of any latency overhead the JIT adds, though.
In terms of real-world scripting, almost any script I've benchmarked has always been ~1.5-2 times faster in Chrome than firefox (versions 4 through 6), except when bound by API calls (last I checked Firefox was on-par or better when dealing with TypedArrays). It looks like TypeInference might be a really big win though.
In terms of real-world scripting, almost any script I've benchmarked has always been ~1.5-2 times faster in Chrome than firefox (versions 4 through 6), except when bound by API calls (last I checked Firefox was on-par or better when dealing with TypedArrays). It looks like TypeInference might be a really big win though.