To test the browsers as fairly as possible we used three brand-new, factory-configured Dell Precision 390 systems with Intel Core 2 Duo chips with 1GB of memory.

All systems ran Windows XP Pro SP2. Each browser was run on its own machine, while the third machine had Internet Explorer 6, Firefox 1.5, and Opera 9 for simple comparison purposes. Performance results were compared on other machines, including a Dell Optiplex GX620 and a Toshiba Portege laptop, and the same relative performance rankings were found giving us assurances that hardware effects did not skew overall placement.

To fully exercise HTML markup, we used the HTML 4.01 test suite from the the World Wide Web Consortium (W3C), as well as test cases derived from the tester's book HTML & XHTML: The Complete Reference.

To test CSS support at a basic level, we ran the W3C CSS1 test suite. To assess more-complex CSS compliance, we ran the W3C CSS2 test suite, as well as more examples provided by CSS guru Eric Meyer.

To assess DOM conformance, we ran the conformance tests posted at www.w3.org/DOM/Test. To assess performance, we measured scripting and rendering speeds.

When talking about JavaScript performance, we ran a variety of tests that included running math function benchmarks, string function benchmarks and simple document.write benchmarks.

To assess rendering performance, we used simple examples of large tables, randomly generated HTML structures in a valid form and other baseline benchmarks. To get a real-world assessment of page rendering times, we downloaded the content of the home pages of popular sites such as MSN, Time magazine, the White House, ESPN and a number of well-known technical and political blogs.