Testcase (details can be found in the comments section):

$ source ~/.java142_09
$ java -fullversion
java full version "1.4.2_09-b05"
$ time java -classpath .:sharedlib/jload.jar DumpTest EXPDUMP > /dev/null
Reading EXPDUMP
End of EXPDUMP reached
finished
31.0u 1.0s 0:41 77% 0+0k 0+0io 0pf+0w

$ source ~/.java142_10
$ time java -classpath .:sharedlib/jload.jar DumpTest EXPDUMP > /dev/null
Reading EXPDUMP
End of EXPDUMP reached
finished
136.0u 1.0s 2:21 96% 0+0k 0+0io 0pf+0w

$ source ~/.java15
$ java -fullversion
java full version "1.5.0_05-b05"
$ time java -classpath .:sharedlib/jload.jar DumpTest EXPDUMP > /dev/null
Reading EXPDUMP
End of EXPDUMP reached
finished
28.0u 1.0s 1:52 25% 0+0k 0+0io 0pf+0w

=> In this example, 1.4.2_10 is ~4.4 times slower than 1.4.2_09 while executing SAPs testcase called DumpTest.