To summarize, we were benchmarking for throughput: how long between the time you click and the time the operation terminates. For a long time, we absolutely crushed Chrome on most of these benchmarks. Firefox started faster, loaded pages faster, switched between tabs faster, JavaScript executed faster, downloads were faster, ...
Chrome developers were benchmarking for responsiveness: how long between the time you click and the user interface shows that something is going on. And that's all that end users actually see. Chrome showed its initial window faster (although it took more time until you could actually do anything with it), was slower at loading pages (but stopped its throbber earlier), slower at switching tabs (but the screen changed faster to show you that it was switching tabs), ...
Which ones were the "correct benchmarks" remains an open question, but it's clear which ones the users perceive/remember as being faster!
Of course, the better architecture that Chrome used meant that they eventually got faster at throughput, too, on computers that had sufficient RAM.
Chrome developers were benchmarking for responsiveness: how long between the time you click and the user interface shows that something is going on. And that's all that end users actually see. Chrome showed its initial window faster (although it took more time until you could actually do anything with it), was slower at loading pages (but stopped its throbber earlier), slower at switching tabs (but the screen changed faster to show you that it was switching tabs), ...
Which ones were the "correct benchmarks" remains an open question, but it's clear which ones the users perceive/remember as being faster!
Of course, the better architecture that Chrome used meant that they eventually got faster at throughput, too, on computers that had sufficient RAM.