Performance degradation is observed in specjbb2005 results measured in HS 19.1-b01 (JDK 6u24 b01) vs. HS 19.0-b09 (JDK 6u23 b05) on linux-amd64 with the following configurations:

-server -XX:+UseConcMarkSweepGC -Xmx1G
-server -XX:+UseConcMarkSweepGC -Xmx6G
-server -XX:+UseConcMarkSweepGC -XX:+CMSIncrementalMode -Xmx1G
-server -XX:+UseConcMarkSweepGC -XX:+CMSIncrementalMode -Xmx6G

HS 19.1-b01 scores are consistently lower by 2-3% than HS 19.0-b09 scores. The scores are approx. 3 standard deviations apart.

Results in Aurora:

http://aurora.russia.sun.com/performance/faces/ChessBoard.xhtml?reportName=J2SEGeneric&parameters=%5Breftags%5D%27ADHOC%27%5Btags%5D%27ADHOC%27%5BrefProductName%5DJava%28TM%29+SE+Runtime+Environment%5Brefrelease%5D1.6.0_23%5Brefbuild%5D%3D+%275%27%5Bbenchmark%5DNOT+LIKE+%27%25.%25%27%5Bshownbenchmarks%5D%28benchmarkresults.name+IS+NOT+NULL%29%5BproductConfig%5D%281%3D1%29%5Brelease%5D%28pr.product.productRelease+%3D+%271.6.0_23%27%29OR%28pr.product.productRelease+%3D+%271.6.0_24%27%29%5Bbuild%5D%28pr.product.build+%3D+%271%27%29OR%28pr.product.build+%3D+%272%27%29OR%28pr.product.build+%3D+%275%27%29%5BrefRunNames%5D%27%27%5BrefBatchNames%5D%27%27%5BrunNames%5D%27%27%5BbatchNames%5D%27%27%5Bvalidity%5Drun.isValid+%3D+true%5Bhosts%5D%28h.machine.name+%3D+%27aurora-x2100-05.russia.sun.com%27%29OR%28h.machine.name+%3D+%27aurora-x2100-06.russia.sun.com%27%29&splitting=%5BX+axis%5Drelease%2C+build%5BComplement%5D%5BZ+axis%5Dos%2C+hostname%5BY+axis%5Dbenchmark%2C+benchmarkConfiguration%2C+productArch%2C+productConfiguration&mixReference=false&viewStyle=chessboard

To get to detailed side-by side comparisons, click on the links displayed as "-x% (-xxx.x)". Example:

http://aurora.russia.sun.com/performance/faces/RunsCompare.xhtml?names=1098.PERFORMANCE_SERVER_PROMOTION-3%2C+1071.PERFORMANCE_SERVER_PROMOTION-3