PerformanceTrying to measure the performance of a web browser is tricky, because of the sheer number of variables involved. With page load times that are often measured in milliseconds, even momentary glitches in network performance can distort results and make test data unreliable. The presence of images, ads, and widgets, often loaded from different servers, can skew results as well. In one test session, for example, I found that pages from a very popular tech site took more than 10 seconds to fully load, regardless of which browser I used, because of issues with a third-party ad server. And then there’s the markup of the page itself, which is often optimized for a particular browser.
I'm sure other publications with more sophisticated test platforms will try to overcome these obstacles and produce definitive benchmarks of IE9 performance. For this post, I can only describe my subjective experience and point to some widely used independent benchmarks.
I can describe the subjective part in one sentence: Holy crap, this browser is fast. For most pages, load times were quick, navigation was snappy, and it was impossible to measure differences between IE9 and rival browsers, even with a stopwatch. A handful of pages in IE9 did exhibit performance problems caused by compatibility issues—I discuss that topic more fully in the wrap-up on the next page.
The single biggest performance boost in IE9 comes from its support for hardware acceleration. Because IE9 runs only on Windows Vista SP2 and Windows 7, it can be tuned to offload some rendering tasks to modern graphics hardware, which often has more raw processing power than the rest of the PC. (Microsoft claims that current browsers use only 10% of a PC’s power, which might be a bit of hyperbole.) It’s clear from daily use, though, that hardware acceleration really does make a difference in rendering text, images, and graphics. As a result, Microsoft finds itself in an unaccustomed position, out in front of other browsers, which are furiously trying to play catch-up. I tested the IE9 beta alongside Firefox 4 beta 5, which was released in September 2010 and is the first Mozilla offering to support hardware acceleration. I also tested it against the most recent beta of Google Chrome 6, which doesn’t use the GPU for rendering. (Google has reportedly placed that feature on its roadmap for Chrome 7.)
The biggest performance differences, not surprisingly, were apparent on Microsoft’s own graphics-oriented tests at its IE Test Drive site. On the FishIE Tank example, which uses the new HTML5 Canvas tag, here’s how the three browsers compared:
IE9’s frame rates stayed high as I kicked up the number of animated fish in the virtual tank. Performance remained smooth and glitch-free even when I moved the window across multiple monitors and docked it to the side of the display using Aero Snap. Firefox 4, by contrast, was able maintain high frame rates for short bursts, but moving the browser frame caused performance to plummet and even froze the display for long periods. Using Firefox, frame rates plummeted dramatically when I selected the most demanding settings (500 and 1000 fish). Chrome, of course, was at unfair disadvantage because of its lack of hardware support. Its frame rates were well below either competitor, although it moved smoothly around the screen without any negative impact on other programs or Windows itself.
For a more independent performance test, I enabled all three browsers for YouTube’s HTML5 channel and tried playing a handful of high-definition videos at 720p and 1080p resolution. All three browsers performed admirably within a window and at full-screen resolution. IE9 and Chrome 6 were able to maintain full-fidelity playback even when tearing a tab out of the browser pane and dragging it to its own window. Firefox 4, on the other hand, failed this test, stopping the playback and starting the clip over when it landed in a new window.
The other new performance-enhancing component in IE9 is the new Chakra JavaScript engine, which uses multiple processor cores and has already been extensively benchmarked via the platform preview releases. Microsoft has handed out several charts showing its impressive improvements; most of those charts include its dismal IE8 score, which is almost an order of magnitude slower than IE9 and grossly distorts the scale of the charts. To make things fairer, I threw out those IE8 results and ran the SunSpider benchmark using only the most recent beta releases of IE9, Firefox, and Chrome. Here’s what the results looked like:
The difference between each browser is only about one-tenth of a second, and that composite result includes dozens of complex operations. The independent
JSBenchmark test produced similar results, with the IE9 beta running 21% faster than the latest Firefox 4 beta but 29% slower than the latest test build of Google Chrome 6. The conclusion? JavaScript performance isn’t a significant differentiator between modern browsers, and IE9 can hold its own with any Webkit-based browser on this score. (Yesterday, in anticipation of this first wave of reviews, the IE Blog published an interesting discussion of
what JavaScript benchmarks really measure.)
In previous IE versions, browser add-ons have been a frequent source of slowdowns. IE8 introduced an add-on manager that was effective but hard to find and too daunting for mere mortals. It’s still available in IE9, but it’s been supplemented with a cleaner tune-up kit. After you run IE9 for a while, you’ll see a notification along the bottom of the screen. If you choose to follow it, you’re taken to this simple list, where you can identify and if necessary disable a troublesome add-on.
Based on my experience, I predict that Google Chrome will continue to win many performance tests. But IE9 has closed the gap impressively.
It’s faster than Firefox across the board and faster than Chrome on tasks where hardware acceleration is involved.
Even when those rivals catch up or pass IE9 (as they probably will), it’s unlikely that performance differences will be significant.
continue on to page 5
