futtta's blog

Firefox 3.1 is just around the corner and I’ve been using the beta’s for a couple of months now, but I didn’t really feel the urge to write about it up until now. But with things heating up between Google Chrome (already out of beta!), Safari and Firefox and with new versions of MS Internet Explorer and Opera in the making as well, one can’t really stay indifferent I guess?

First off; a non-exhaustive list of changes;

• support for <audio> and <video> elements (cfr. html5 draft specs), with ogg/vorbis, wav and ogg/theora compiled in
• lots of css-improvements
• Javascript performance benefits hugely from the new Tracemonkey-engine (just ran Sunspider and got a score of 2482.2ms, compared to approx. 5000ms in FF3 and 20000ms in FF2). Performance could also benefit from webworkers, which allow javascript-webdevelopers to background the execution of javascript.
• javascript-hooks to allow sites to do geolocation
• private browsing (porn-mode, finally!)
• and lots of smaller changes, such as selective tab restore after crashing (so you can kill the culprit which crashed your session), the possibility to move a tab to a new window, support for the “defer”-attribute in the script element, clickable hyperlinks in the source view, …

So if FF3.1 performs that great in Sunspider, does it really feel that much faster as well? To be honest; it doesn’t. Or at least, it didn’t, at first. But here’s a tip; if you’re a bit like me you’re bound to have a lot of extensions installed (and disabled and uninstalled and not compatible and …), you might have some forgotten tweak in your about:config and you probably have huge history and bookmark-databases. In that case do yourself a favor and start from scratch with a new profile and Firefox 3.1 will truly fly.

Off course not all is perfect. I don’t like the fact that tabs inadvertently get moved to a window of their own regularly. And Flash still crashes FF all too often, Firefox really needs something like the process isolation in Google Chrome and MS IE8′s loosely coupled IE, but that might be more than just a small CR.

All in all, with Firefox 3.1 the Mozilla-folks seem to have almost everything to fight the new kid in town. You can download the latest beta here and test for yourself. Let those browser-wars rage!

As the official release of Firefox 3 is getting closer, with Release Candidate 1 being available since May 17th, I decided to boldly go where codinghorror has gone before and do a quick-and-dirty Javascript-performance comparison of the different browsers I’ve got installed on my Dell Latitude D620 laptop, using Webkit’s Sunspider benchmark.

Let’s start with the results for the browsers on my Windows XP SP2 installation, ordered from slowest to fastest. Each test was executed 2 times, clicking on the results will teleport you to the detailed results where you can paste the URL’s of another test to compare.

• msie 6 (6.0.2900.2180.xpsp_sp2_qfe.070227-2300): 47203.0ms +/- 16.7% and 47882.2ms +/- 4.9%
• IE 7 (standalone version, see below): 44726.2ms +/- 4.1% and 42655.2ms +/- 5.0%
• firefox 2 (2.0.0.12): 26025.4ms +/- 4.5% and 25305.0ms +/- 1.1%
• opera 9.27: 14202.0ms +/- 1.0% and 14755.6ms +/- 2.4%
• opera 9.5 (b2, build 9945) produced a number of ‘NaN’s’, but is clearly faster then it’s predecessor
• safari 3.1 (build 525.13): 6759.0ms +/- 1.2% and 6750.8ms +/- 2.0%
• firefox 3 (rc1): 5830.4ms +/- 2.2% and 5765.8ms +/- 1.0%

The MSIE7-results are probably not entirely representative, as I use Tredosoft’s standalone IE7. This is a bit of a hack to have IE7 on my otherwise MSIE6-based system. Moreover my corporate Windows-installation is infested with crapware, notably McAfee OAS and Zonealarm seem to be slowing things down enormously. The codinghorror-tests indeed show significantly better results for this browser, although IE does have serious issues with string concatenation, which should be fixed in IE8.

On the same hardware, but booting in Ubuntu 8.04 (Linux) form my external USB HD (a.k.a. my ‘disktop‘), I got the following results:

• opera 9.27: 15343.2ms +/- 1.1% and 15499.4ms +/- 1.1%
• firefox 3 (rc1, official mozilla build): 5352.6ms +/- 1.1% and 5343.8ms +/- 0.6%
  
• firefox 3 (b5, included in ubuntu 8.04): 5195.2ms +/- 1.6% and 5240.2ms +/- 1.4%
• konqueror 4: not tested yet, results will follow later today can’t get test to completely run, any KDE-user want to give this a try?

Firefox 3 RC1 seems slightly slower then b5, but maybe the Ubuntu-b5-version is compiled with optimizations? Firefox is also faster on Ubuntu, but the anti-virus-bloat is probably messing with our heads here (although Opera is slower on Linux, go figure).

The general conclusion however; Firefox 3 is a huge step forward as far Javascript-performance is concerned. Users of javascript-heavy web-applications such as Gmail, Google Reader, Zoho Office and Zimbra should benefit enormously from this. It would however be very interesting to perform similar tests with regards to ‘normal page rendering’ (html/css). Does anyone know of such benchmarks?