The competition for performance proficiency among recent browsers has doubled. All are used daily and they have all made significant efforts to improve their speed and security. There are now new tools to analyze, in-depth, the performance of each browser.
In order to determine the best browser software, we’ll compare the performance of the Internet’s main web browsers. This article will cover Internet Explorer 9, Firefox 4, Google Chrome 10, Safari 5.0.4, and Opera 11.10.
There are no subjective points in this article. We have only included measurements, tests, and comparisons.
<bold>N.B. Test setup: Intel Core 2 Duo E7500, with 2 GB of DDR3 and a small Radeon HD 3450 graphics ATI card. It is important to specify this information, because it influences the performance.
Differences from one PC to another may exist, particularly in terms of graphical acceleration. We used the Windows 7 operating system. We also did a test without a module installed.
Table of Contents
Performance
In terms of performance, we performed the Javascript tests developed by Mozilla and Google’s V8 Benchmark, Dromaeo DOM and Kraken. We also performed other Javascript tests, and Peacekeeper takes DOM operations into account so that nothing is forgotten.
Let’s start with Peacekeeper, a performance test developed by Futuremark. It is a very complete Javascript performance test. The final browser score is the addition of multiple tests to see how it performs in terms of rendering graphics, including complex returns with HTML5 with physical calculations (rendering and Complex Graphic).
The Social Network grade measures the loading functions of a classic page, a search, and a ranking. Data is another important grade, which measures the capacity on a dynamic page with javascript zones to add, remove, or modify elements (e.g. text or Facebook images). Text Parsing specifically measures text, such as applying language filters or scaling:
Google Chrome 10 surpasses the other browsers due to its Data grade, although it is not the strongest in other areas. Opera 11, for example, has the weakest Data grade but is first in terms of graphical returns, DOM applications, text, and all other criteria.
We also used Sunspider 0.9.1, which is a very relevant tool that measures mathematical functions, 3D, and data encryption capacity. It gives good insight into how fast a browser is without Flash or HTML5.
This would explain the particularly good score of 216 ms for IE9. Google Chrome 10, Firefox 4, and Opera 11 follow, with 247, 252, and 260 ms, respectively, Safari 5 trails behind the group, with 323 ms.
Browsers are not static across the board on HTML5. IE9 and Firefox 4 can, in certain places, use the graphic card to increase their performance. In all comparative studies (i.e. the GUIMark2 test for fluency in NeverMind the Bullets), we generally find IE9 ahead of Firefox 4, then side-by-side are Opera 11 and Google Chrome 10, followed by Safari 5, which needs an update.
Flash technology can be subjected to a test with FlashBenchmark 8, which measures particularly the performance in flash games. Now, it’s Safari 5 that is the best, ahead of Opera and IE9, followed by Firefox 4 and Google Chrome 10.
Another very revealing test is the performance of browsers in CSS2.1 style and CSS3.0. It measures the browser’s ability to feel comfortable between style sheets and complex HTML collations. Here, we were required to violate our own rule and use a test provided by Microsoft because there is no other relevant one.
The browser should try to get out of a complex maze (i.e. 40 x 40 for this test) as quickly as possible. Opera 11 clearly dominates, with 16 seconds. IE9 continues, with 32 seconds, Safari 5 is in the third place, needing 221 seconds to exit. Firefox 4 required much more patient, taking 809 seconds, followed only by Google Chrome 10’s 869 seconds.
| Ranking of browsers according to their performance | ||||||
| Navegador | Peacekeeper | Sunspider | HTML5 | Flash | CSS2.1/3 | Total |
| Opera 11 | 2 | 4 | 3 | 2 | 1 | Second |
| IE9 | 3 | 1 | 1 | 3 | 2 | First |
| Chrome 10 | 1 | 2 | 4 | 5 | 5 | Third |
| Firefox 4 | 5 | 3 | 2 | 4 | 4 | Fourth |
| Safari 5 | 4 | 5 | 5 | 1 | 3 | Fifth |
Resource Management
For this criteria, we are looking to measure each browser’s memory load and processor use peaks. With several tabs open, and after several minutes of use, can the browser be adjusted, or does it become increasingly greedy? We decided to use the Linternaute.com home page, which is animated with advertising and flash panels, as our reference page. Our findings were, then, confirmed in a test using Youtube, with 10 video tabs open for 30 minutes.
| Memory load in MB and processor utilization in % |
|||
| Browser | linternaute.com | 20 tabs from linternaute.com | 30 minutes after |
| Opera 11 | 96 MB / 0 a 6% | 318 MB / 2 a 15 % | 270 MB / 2 a 25 % |
| IE9 | 70 MB / 0 a 3% | 700 MB / 20 a 40 % | 500 MB / 10 a 30 % |
| Chrome 10 | 56 MB / 0 a 4% | 290 MB / 0 a 8 % | De 225 a 350 MB / 2 a 40 % |
| Firefox 4 | 59 MB / 1% | 290 MB / 10 % | 400 MB / 10 a 40 % |
| Safari 5 | 56 MB / 0 a 10 % | 215 MB / 2 a 35 % | 310 MB / 8 a 55 % |
N.B. It is noteworthy that we privilege the use over time and a resource load that remains reasonable and stable, and even decreases over time while using the same sites.
Firefox 4 processes flash content and silverlight content separately. The base memory load of Opera 11 is explained by its rich base (i.e. email client and integrated ftp).
Opera 11 excels in this field. Google Chrome 10 fares pretty well well, with a regular refresh of the memory used. IE 9 consumes quite a lot of resources, but it reduces and stabilizes over time. Firefox 4 and Safari 5 see their memory load increase over time, with spikes in processor activity occurring in the case of the Apple browser. They are the only two browsers that must be started again on a regular basis if you do not want to choke the PC.
Web Standards
In regard to web standards great progress has been made, mainly by Microsoft. In the ACID tests, IE9, with 95/100, and Firefox 4, with 97, just nearly reach the maximum garde. The same goes for the CSS selector, which defines the format of the content: all have the maximum grade, except Google Chrome 10, which is lacking in a few points.
Sputnik tests browsers and grades them with a maximum of 5,246 points. The lowest score is for Firefox 4, with 5,065 tests approved, and the highest score is for IE9, with 5,174. The results of the 5 browsers are globally satisfactory, but they can be improved.
For HTML5, there really is no official test, since neither the HTML5 Working Group nor W3C (World Wide Web Consortium) reported on what the standards of this language will be. However, the HTML5 site proposes several tests on browser compatibility with HTML5-enabled features, such as video, audio, page layout, or applications.
Google Chrome 10 dominates with a score of 288 over 400, followed by Opera 11 (258), Firefox 4 (240), Safari 5 (187), and IE9 (130). This grade is in reference to the browser’s respect for the standards and not the HTML 5 performance, as previously mentioned.
IE9, which was the best in HTML 5 renderings, does not seem to be the best browser to apply the first few recommendations of HTML 5 uses.
Security
In terms of browser security, great strides have been made, namely in regard to the reaction time of the editors to correct an uncovered fault. In 2009, Apple took an average of 13 days to correct a Safari bug, with peaks of up to 40 days. Today, publishers are generally delayed 24 hours, with peaks at Microsoft at a 4-day delay before correcting a vulnerability. Apple may take up to 3 weeks to resolve an SSL certificate problem to secure Internet transactions.
Reaction time is the most important. But how many failures do we find each year? This depends on the browser. Symantec informs us in its annual balance that it is in Google Chrome where we can find the most failures. However, the reactivity of Google attenuates this weakness.
This is the graph of Symantec’s Internet Security Threat Report 16 annual report, which shows the number of failures identified in each browser in 2010 and 2009.
We have ignored the measure of “protection of private life,” because we found it impossible to isolate the advantages and faults of the browsers and the elements that depend on websites, search engines, cookies, and user habits. In other words, we can’t find a way to test it in a relevant way.
Conclusion
Opera 11 has a margin of advantage according to us. It is very good in terms of overall performance and behaves very well in resource management. It is a browser that is always respectful of web standards and is also very secure, with few, rapidly-corrected flaws discovered.
IE9 is also good, and it is the browser that starts the fastest. It is very comfortable for daily use. Google Chrome is also fully usable.
Firefox 4 is a slight disappointment, since it has a demanding side that can monopolize up to 1 GB of RAM. As for Safari 5, it is a browser that is currently outdated in Windows.
