I've added simple tracking of page loading time on one of the sites I maintain.
var starttime = (new Date()).getTime();
…
ga._trackEvent('loading-speed', 'time', 'onload', (new Date()).getTime()-starttime);220,000 pageviews were logged via Google Analytics Event Tracking API (1000 pageviews from iPad and 100 via IE Chrome Frame). Loading times over 20s have been excluded. Results don't include mobile phone browsers, because these were sent away to suffer a „mobile-optimized” site.
Sorry, I don't have standard deviation data
Results are average loading times across all pages of the site. I didn't break down stats by subpages or cache state, because I wanted overall picture with real site usage patterns.
Hugs and kisses for Inkscape for converting graph to SVG
Graph is sorted by time needed to load whole DOM, because that's when the site becomes usable. „Half of page” is an arbitrary point before majority of JavaScript is loaded. Fully loaded is time between script ran in <head> to onload event.
Please remember this isn't indicative of browser speed! There are other factors correlated with browser usage that contribute to measured speed.
- I guess Linux browsers get worse times because mainstream users use them only on netbooks.
- iPad Safari doesn't have on-disk cache — only tiny RAM cache like iPhone — which forces it to reload pages, even on Back/Forward navigation.
- Google Analytics reports most iPad users have „DSL” connection and most popular networks belong to landline broadband providers, so I suspect 3G iPads are not the reason why iPad gets worst loading times.
