Firefox 3.6 Beta 1, like every other browser, makes a claim to being "faster." We took Firefox and all the other latest browsers, put them on Windows 7, and ran them through our human-measured speed tests to vet the bragging.
We've done a good number of these tests now, and the methodology remains much the same here—testing how long it takes for browsers to start up and load pages, and how much memory is eaten up, from a user's perspective. We don't use a fancy multi-protocol benchmarking suite, mostly because each suite is subjective to a developers' preferences and recording errors.
Browser start-ups are measured from double-clicking to the load of a locally saved Google home page, as "cold," or right after reboot and "warm," with the browser already having run once. Each browser is given a folder full of nine sites—up from eight in previous tests—and forced to load them all at once. Those timings are measured with Rob Keir's timer app, and done three times each and averaged, with way-out results discarded under the assumption of general computer wonkiness.