Peacekeeper tests under VirtualBox

512MiB RAM, 32MiB vid, 3D hw accel, single core, hardware virtualisation enabled

Results from March 2010

BTW, I've given up on testing on Peacekeeper. There are some serious flaws in their benchmark that they have not corrected. Basic errors in the harness, failure to measure what they think they are measuring (what is being measured is the test harness or due to badly presented problem, how well the browser can hardcode a hack workaround [chrome has a hardcode for the silly .* in the peacekeeper regex for example, other browsers have yet to add the hack]). Last I checked though, on physical machine (not even running Vista, so no D2D) Opera was on top at 6220, followed by Chrome 14 at 5267, follow by Firefox 7 at 4349 and Firefox 5 at 3688. IE8 was 675

Identical VM image used in order to test IE7/IE8/IE9 on this hardware. A move to Vista was forced in order to test IE9. IE9 useragent forcibly changed in registry since it reports IE8 to peacekeeper otherwise. FF3.7 is currently suffering from a regression due to changing regex caching due to new javascript requirements. Also seems to have taken a slight hit in DOM about 5-6%. IE9 while doing the array/string tests did not display the splash image, but that could just be a bug in it. It did not run Canvas, but I understand it is not yet capable of it. This does not count against it in the Peacekeeper benchmark. Detail images are not yet updated. If peacekeeper had a private URL for benchmarking, could just link people to the peacekeeper page.

Some people have raised objections to my testing on a virtual machine, even on a host with hardware virtualisation. Below the virtualised set are my results on a non-virtual machine. IE8/9 are excluded, IE7 was run using IE7 standalone. The gap between Minefield and Chromium does shrinks from 37% to 20% and Opera and Firefox 3.5 are neck and neck. Of course this is on a completely different set of system specs.