- Startup time, both cold (after a PC reboot) and warm (after the browser has already been running)
- Hardware acceleration tests
- Independent studies of page loading performance
Maxthon. This outlier from China has surprised many with its extensive support for HTML5 and performance on benchmark tests. It can display pages using either Chrome’s Webkit page renderer or IE’s Trident. Maxthon dolls up the browser with many tools for things like finding and downloading all the media on a page. It even includes a Night Mode that darkens primarily white Web pages to save your eyes.
Microsoft Internet Explorer 9. For this time, we’ll stick with what most people are using, but we’ll be sure to test Microsoft’s latest, Internet Explorer 10 on Windows 8, in a future roundup. Yes, IE10 is available for Windows 7, but only as a Release Preview at this point.
Opera. When it first hit way back in 1996, speed was the main hallmark of Opera. It was notably faster than the competition at the time, but the only problem was that it didn’t display all websites correctly. Opera has also been a major innovator. It was the first browser with things we take for granted in all browsers today, including such basics as tabs and built-in search. And the innovation hasn’t stopped: Opera recently became the first to implement standards-based access to the webcam and support for paged content.
Astute readers will notice that I’ve left out Safari. Apple’s browser for Windows is no longer promoted on the company’s Safari download page and has not been released in new versions as Safari on the Mac has. further ado, let’s start with the results. First up: startup time.
One of the most frustrating effects of slow performance in any software is when you have to wait just to start using it. This delay is far more pronounced the first time you start running a browser after booting up the OS—this is called a “cold” startup. Once you’ve run an application and closed it, the next time you start it is much faster—a warm startup. Startup time has been a bugaboo for Firefox for a number of years, but things have gotten better for Mozilla’s browser recently in this department.
|Browser||Cold Startup Time
|Warm Startup Time
|Internet Explorer 9||3.0||1.5|
Occasionally any browser will take much longer to cold start, if it’s updating or loading new extensions, for example. As you can see, our old friend Opera is the biggest offender this time around. But after the initial run, in warm startup all the browsers are acceptably fast. Firefox is no longer trailing by much, and Chrome is expectedly fast. But this first round goes to Internet Explorer 9.
All of these benchmarks claim to test real-world browser operations, and each of these benchmarks consists of a set of server operations run multiple times. SunSpider’s page, for example, states that the benchmark “generates a tagcloud from JSON input, a 3D raytracer, cryptography tests, code decompression, and many more. But in the end, they’re all really synthetic tests that only test one slice of browser performance. Here were my results for the software under consideration:
|Browser||SunSpider 0.9.1 Score in ms
(lower is better)
|Google Chrome 23||236|
|Internet Explorer 9||260|
Chrome has regained the lead on SunSpider, after trading places with IE and Opera temporarily. But the largest difference is only by a margin of 66ms, where in the past it had been in the thousands.
|Browser||Google V8 (v.7) Score
(higher is better)
|Maxthon 3.4||9767||Google Chrome 23||9694|
|Internet Explorer 9||2048|
This one was a bit of a shocker to me, that the little-known Maxthon could beat Google’s browser in the company’s own benchmark. But the two top performers here, Chrome and Maxthon, are separated by a far smaller margin than that between them and the rest. IE9 makes a particularly poor showing on this one. For a future article that includes IE10, I’ll use Google’s newer Octane benchmark suite, which is a superset of V8. But for now, the newer test isn’t compatible with IE9.
|Browser||Mozilla Kraken 1.1 Score in ms
(lower is better)
|Google Chrome 23||3235|
|Internet Explorer 9||16794|
The developers of Internet Explorer 9 introduced a number of browser innovations, but perhaps the most significant of these was hardware acceleration. This technique speeds up all kinds of browser functions using the PC’s graphics hardware, and is particularly important for 3D visualizations like those used in some online games. The other browsers, starting with Firefox, have followed suit, implementing hardware acceleration in their code. Notably, with Firefox and Chrome, you get the speed boost not only in Windows 7 and 8 as is the case with IE, but in Mac OS and Linux, too.
The IE team has posted a raft of demos that test this performance at IETestdrive.com. I use two of these to compare hardware acceleration. I also use a different test PC with this group, one with a beefier graphics card. The first test, Psychedelic Browsing, spins a color wheel and plays spacy sounds, reporting RPM as a result. The test clearly separates browser with hardware acceleration from those without. Opera hasn’t yet implemented hardware acceleration by default, and Maxthon, which up to now has looked great, shows its Achilles heel. Here are my results, using a 3.4GHz quad-core desktop with an ATI Radeon HD4290 graphics card:
RPM (higher is better)
|Google Chrome 23||3390 (correct sound)|
|Internet Explorer 9||3303 (correct sound)|
|Firefox 17||3157 (no sound)|
|Opera 12||782 (no sound)|
|Maxthon 3.4||57 (correct sound)|
The second IETestdrive benchmark I use is HTML5 Particle Acceleration. This benchmark spins a globe composed of small spheres and reports Draw Time (milliseconds- lower is better) and FPS (higher is better). FPS tops out at 60, the maximum for a standard LCD monitor. The test also reports a Score, but that fluctuates too drastically to accurately report. Below are my results for the current crop of browsers on the 3.4GHz quad-core desktop with an ATI Radeon HD4290 graphics card:
|HTML5 Particle Acceleration Benchmark||Draw Time
(milliseconds- lower is better)
|FPS (higher is better)|
|Internet Explorer 9||6||60|
Chalk one up for Firefox! Even though this test seemed a little unfair to Microsoft, which has designed the test for IE10 (which gets TKms and TKFPS), IE9 acquits itself with aplomb. Again, we see that Maxthon and Opera lag in hardware acceleration.
One final test of hardware acceleration (which comes from Mozilla): the Hardware Acceleration Stress test, which spins a spiral of photos in the browser window and reports a score in frames per second. This test showed the browsers furthest along in implementing hardware acceleration (particularly for CSS) to good advantage. Note the benchmark doesn’t report frame rates over 60FPS, since that’s the limit of standard LCDs. Because of that, it’s pretty much a pass/fail test, where anything under 60 is a fail:
|Browser on 3.4GHz quad core with ATI Radeon HD 4290||Mozilla Hardware Acceleration Stress Test
FPS (higher is better)
|Internet Explorer 9||60+|
|Google Chrome 23||60+|
Again, this hardware acceleration benchmark reveals the losers more clearly than the winners. Firefox, IE9, and Chrome all pass with flying colors, while Opera and Maxthon trail. With a three-way tie in this one, and the previous test results too close between the top three finishers, I have to declare a draw between Chrome, Firefox, and IE9 on this one.
Page Load Reports
While benchmarks show one side of browser speed, performance measuring services get to see literally billions actual page load times for actual websites. One such organization is New Relic, which periodically publishes reports of how fast each browser loads pages it’s monitoring. The latest report came out just this past November, and here’s how the browsers pan out running in Windows:
Surprisingly, IE9 and Firefox both beat out the perceived speed champion, Google Chrome, though by only a fraction of a second. One concerning thing to note here is that the more recent version of Chrome, 22, came out slower than the earlier version 21, by a wider gap than that between version 22 and the competing browsers.
Another page load performance measurement comes from Strangeloop Networks, which provides front-end optimization (FEO) solutions to global ecommerce websites. In a February test, the firm used the WebPageTest tool to measure performance of the Alexa-ranked top 2,000 e-commerce sites. Here’s a summary of their findings:
StrangeLoop found that IE9 and Firefox 7 were each about 5% faster than Chrome in this study with IE9 winning by a tiny margin of .38ms. So IE9 takes this round.
The Fastest Browser
So Here are the winners for each category:
Startup: Internet Explorer
Hardware Acceleration: Three-way Tie: Chrome, Firefox, Internet Explorer
Page Load Reports: Internet Explorer
Overall Winner: Internet Explorer
Yes, it will be a shocker to many, but this isn’t your grandpa’s Internet Explorer. The team behind Microsoft’s browser has been hard at work making the much-derided browser more “performant.” But I hope that the criteria above show that speed can be measured in multiple ways. There are all sorts of ways you could weight the results to get a different answer, depending on what you think is the most important component of speed. In any case it’s a very tight race between the top three, with IE by the smallest of noses.
Remember, also, that speed isn’t the only thing you should base your browser decision on: Support for standards like HTML5 and CSS3 are equally important—in fact, one standard supported by Chrome and Firefox, SPDY, actually can mean much faster performance for certain sites. And don’t forget features like tab implementation, syncing, security, and extensibility. For in-depth looks at how each browser performs not only in speed, but in these other important criteria, dig into our full reviews of Web browsers.
By Michael Muchmore, PCMag