Comparing Ruby1.9 and MacRuby (another JITted Ruby) on my MBA (which is slightly slower than the other machines used here):

ruby1.9 ~/tmp/ben1.rb11.44724430.074118macruby ~/tmp/ben1.rb4.6390053.818616Obviously the unboxed floating point values of MacRuby seriously win, but also the JIT seems to win (but not as much as V8, unsurprisingly).V8 version 2.2.4.2> function fact (n) { return 1 a = new Date; for (var i = 0;  i a = new Date; for (var i = 0;  i  (these are milliseconds).

MRuby (0.0.0):

avocirb>  def fact n; (n > 1) ? n*fact(n-1) : 1; end
avocirb> $a = time; for i in 1 .. 10000000; fact(10.0); end; p time – $a
57.8161611557007
avocirb> $a = time; for i in 1 .. 10000000; $b = fact(10.0); end; p time – $a
59.1052560806274
avocirb> p $b
3628800.0

Javascript (V8 3.9.4):

avocado> function fact (n) { return 1 a = internal.time(); for (var i = 0;  i a = internal.time(); for (var i = 0;  i print(b);
3628800

The results are in line with what you measured. V8 was about the same magnitude faster than MRuby as in your test.
For curiosity I also cross checked the results against PHP.

PHP (5.3.6):

cli> php -r ‘function fact($n) { return ($n > 1) ? $n * fact($n – 1) : $n; } $a = microtime(true); for ($i = 1; $i php -r ‘function fact($n) { return ($n > 1) ? $n * fact($n – 1) : $n; } $a = microtime(true); for ($i = 1; $i < 10000000; $i++) { $b = fact(10.0); } print microtime(true) – $a; print "n"; print $b; print "n";'
36.732476949692
3628800

So MRuby and PHP execution times at least have the same order of magnitude, which was what I expected because I think they have the same code execution model.