We performed all of our testing on a Gateway P-7808U notebook running a clean installation of Windows Vista Service Pack 1; we reinstalled the operating system before testing each browser. For each browser, we cleared the browser's cache and then loaded each page in our test suite. We repeated the process ten times per site per browser to ensure accurate results, to factor out fluctuations in network traffic, and to build a sufficiently large sample size to identify trends. In addition, we threw out the two best and the two worst scores for each page load test to further reduce the influence of fluctuations and to produce more consistent results.
Some browsers will report that a page has finished loading even though parts of the page haven't yet appeared. We didn't base our determination of when a page was loaded on the browser's opinion. Instead, we relied on whether all visual elements of the page were loaded and ready to use. For example, on Yahoo's home page, we judged the page to be ready when all of its graphics and images were loaded.
A Grain of Salt Required
Measuring browser performance is not an exact science. Many, many variables are in play, such as a site's server load at the time of testing, when you consider a page to have finished loading, human error, and so forth. Testing methodologies vary as well, and range from the browser benchmarks mentioned above to stopwatch testing. Each testing methodology could yield dramatically different results from others.
Still, at least in our tests, Chrome and Firefox 3.5 impressed us with their overall snappy feel. And, of course, speed isn't all that matters in a browser; each browser we tested is certainly "fast enough," and each has its own special set of features, so if you're looking on switching to a different browser, take a look at all of them, and go with whichever one best fits your needs.
Sign up for Computerworld eNewsletters.