Internet Explorer 10 Tops New ‘Robohornet’ Speed Test
Want to stress-test your browser of choice? Head on over the Robohornet site, but be forewarned its a long test and there’s a good chance your browser is going to fall on its face — unless of course you’re using Internet Explorer 10.
Robohornet was created by Google’s Alex Komoroske, but it’s an open source project with “stewardship committee members” that extend well beyond Google. There are representatives from Facebook, SmugMug and Sencha participating, as well as individuals like Tom Robinson, creator of the Cappuccino framework.
Robohornet is somewhat novel in that it’s trying to be a community-driven benchmark. The tests that comprise the benchmark can be created by anyone. What’s more, even web developers not interested in writing tests can still participate by voting on which tests to include. See the Robohornet GitHub page for details on participating.
Of course, while Robohornet sounds really good up until right here, we have some bad news for you — right now most of the tests are what’s known as microbenchmarks, very small, highly abstracted tests that often have very little bearing on real-world performance.
As Mark Twain once said, there are lies, damn lies and browser benchmarks (or words to that effect) and in its current state Robohornet may well be doing more harm than good.
Microsoft has already come out dismissing Robohornet as not “representative of the performance users might encounter on real-world sites.” What’s most interesting about Microsoft’s reaction is that Internet Explorer 10 on Windows 8 actually tops the Robohornet tests, besting Chrome, Firefox, Safari and Opera according to some early tests by Tom’s Hardware.
Not happy with the Robohornet tests, Microsoft has created its own Robohornet-derived benchmark suite it calls Robohornet Pro, which the company claims better represents “real world” sites.
Mozilla’s Justin Lebar and Nicholas Nethercote have also both been critical of Robohornet, filing a bug entitled “eliminate and outlaw microbenchmarks.” “If you guys want us (in my case, Mozilla) to take Robohornet seriously,” writes Lebar, “I strongly recommend you write some macrobenchmarks and eliminate the microbenchmarks from your test suite.”
Lebar goes on to say that he really likes “the idea of a community-driven benchmark. I hope that aspect of this project, rather than the microbenchmarks, becomes the hallmark of Robohornet.” Would that it were so.