17 May 2010

Unofficial browser speed tests (Round Two)

About three months ago I made a series of tests on my computer to check how fast the browsers of the day really were. Of course, three months in the online world is a very long time: has recently moved up to version 6, at least in the dev channel; this is usually an indication that the stable channel will soon follow; managed to finish version 10.50 just in time to be added to the browser choice screen, even though that didn’t really change the hierarchy on the browser market; and Internet Explorer tries to get back in the game with his next version, currently available as a ‘Platform Preview’. Besides, I have also changed my PC in the mean time, so the results from last time are clearly out of date.

So I decided to repeat the tests with more-or-less the same conditions as before: the online test suites SunSpider and V8. For V8 I ran the test three times in each browser and averaged the result; they tend to vary quite a lot from a run to the next. SunSpider already delivers an average of 5 test runs. The browsers are also a little different: I used the most recent stable versions of Chrome, Firefox, Opera and Internet Explorer, as well as the alpha version of and Internet Explorer’s Platform Preview #2.

This time the results are closer to the conclusions of other online tests and less surprising. Both Chrome and Opera are in first place, Chrome leading in SunSpider and Opera, by a smaller margin, in V8. Internet Explorer 8 is the last in both cases by a huge margin, while it’s preview version competes for third place with Firefox 3.7 alpha. It’s pretty remarkable how much Internet Explorer has improved in terms of JavaScript performance in these two previews, becoming approx. 7.5 times faster than the current release and beating both Firefox incarnations, at least according to SunSpider. On the other hand, it looks like Firefox did not make a lot of speed improvements in this alpha, as the team is concentrating on the integration of multi-process architecture and standards. So, on the final release day, will we see for the first time an Internet Explorer version faster than Firefox?! Browser test results SunSpider

The is one result though that raises some questions, although it’s not on these charts: I ran the tests with Chrome incognito and found, to my surprise, that SunSpider reports a much lower result than the non-incognito version, somewhere around 865ms! That would make it slower that Opera and closer to the Internet Explorer preview. The phenomenon does not occur with the V8 suite however, which gives Chrome incognito a score of about 2600, comparable with the standard browser. It’s hard to imagine one of the features that gives Chrome it’s speed is disabled in incognito mode, unless it somehow relies on the cache. I think it’s more likely SunSpider has some sort of glitch triggered by the incognito more. After all, no test can be perfect and every result should ideally be backed up by another test suite.  Browser test results V8

It’s interesting how the browsers with the smallest market share are the most advanced in terms of speed. After all, it’s only natural that the challengers offer something better to attract users from competitors, especially from the bigger ones. On the other hand, I’m not sure this speed race will go on for much longer. Using Firefox alpha during these tests and after that, as briefly as it was, I didn’t feel a noticeable slowness compared to Chrome. If users don’t feel the performance increase in their everyday browsing, sooner or later the competition will focus on other areas. Or maybe the online applications will become more and more powerful, driving move innovation on speed, like Google’s initiative Native Client.

Post a Comment