Browser war. Summer 2010

Published on August 05, 2010

Browser war. Summer 2010

    I want to say right away that this testing will be carried out exclusively for research purposes. There are no goals to identify the best browser. The data will be presented to familiarize yourself with the current trends in the browser market.

    The browser war began back in 1996 with a fierce confrontation between Netscape Navigator and Internet Explorer. You can read more about this on Wikipedia . I don’t want to go into a historical digression and lyrical digressions, therefore I will proceed to the most comparative testing.

    Who will we compare? List of experimental rabbits:
    one). Opera 10.60.3445
    2). Google Chrome 5.0.375.125
    3). Google Chrome 6.0.484 Canary
    4). Mozilla Firefox 3.6.8
    5). Mozilla Firefox 4 beta 2
    6). Internet Explorer 8
    7). Internet Explorer 9 PP4
    8). Safari 5.0.1
    9). Maxthon 3.0.15.300 RC
    10). Opera 10.70.3468

    In some tests (JavaScript performance), configuration will be very important. Here she is:
    Intel Core 2 Duo P8400 2.26 GHz
    2 Gb RAM
    3072 Cache L2
    NVIDIA GF 9600M GT 512 Mb
    MS Windows 7 Home

    Why Windows 7 and not Linux? Because some tests were written for browsers under Windows (for example, Futuremark Peacekeeper), so on my Ubuntu browsers give much less results in a synthetic test. However, the difference between browsers in the same JavaScript processing remains (as a rule). So browsers that have their own versions for Linux (Google Chrome, Mozilla Firefox, Opera) show about the same difference in results. Why not other OS? Because I don’t have them in my cars.

    Support for web standards and all that



    Acid3

    Taschemta, let's start with the browser support for the W3C specifications. The most common web standards test is the Acid3 ( Test Information ) test . I note that the test versions of browsers that have passed the test in stable branches will not be tested, because it is an empty lesson. Sometimes in some preliminary assemblies there are regressions in terms of passing some tests, but they fix everything to stable branches.
    So, let's go:

    Opera 10.60.3445

    image
    Opera 10.60 passes the test at 100 out of 100, but it does not show very satisfactory results in terms of speed of passing the test, although it is quite acceptable. I want to note that the Acid3 test is not very suitable for measuring performance, so you should not focus on the speed of passing. So Opera 10.70 loses the honor of being tested

    Chrome 5.0.375.125 Stable

    image
    The test for specifications passes with a bang, there are troubles in the speed of passing the test, but in the case of Acid3, I repeat, this is not critical. Canary will also not be tested on Acid3.

    Firefox 3.6.8

    image
    The popular creation of the Mozilla Foundation could not pass the test completely, failing 6 subtests. Therefore, you will have to run the test on the beta version of Firefox 4.

    Firefox 4.0 b2

    image
    Progress in the test version is observed, although not so obvious. The beta version was able to pass three more tests, having passed 97 out of 100 subtests.

    Internet Explorer 8

    image
    Microsoft's browser definitely cannot handle 80 subtests. Very poor result, and even with noticeable brakes.

    Internet Explorer 9 PP4

    image
    The fourth test version of Internet Explorer 9 made a good leap compared to IE 8. The browser developers were able to “teach” the web browser another 75 subtests, which is very pleasing.

    Safari 5.0.1

    image
    Safari 5.0.1 completely passes Acid3.

    Maxthon 3.0.15.300 RC

    image
    In the test, I decided to include the curious development of Chinese programmers - Maxthon. It used to be known as MyIE, which was essentially an add-on to Internet Explorer, diluting it with a number of interesting features, but the "heart" of MyIE / Maxthon was the Trident rendering engine. The developers got tired of it and, starting from the 3rd version, got divorced from Trident and married WebKit, although at the same time it is still possible to use Trident (this is claimed as a mega cool feature). In general, I finish the excursion. Acid3 test "numerically" the browser passes, but the image is somewhat crooked that kakbe hints that everything is not right in the Danish kingdom. But I think that it’s possible to forgive the developers, because so far there hasn’t been an official release and there is still time to fix bugs.

    CSS3

    This fun thing for the design of web pages, although not yet fully developed, is already used by the creators of web pages. Some specifications have already been approved by the Internet consortium as recommendations, so the CSS3 Selectors Test was developed . So, go ahead:

    Opera 10.60.3445


    Opera 10.60 all 574 tests, which is good news. Opera 10.70 will not be tested.

    Chrome 5.0.375.125 Stable


    All tests passed with a bang. The canary will not be tested again.

    Firefox 3.6.8

    image
    Mozilla Firefox is rehabilitating for minor setbacks in Acid3 by passing the full CSS3 Selectors Test. There is no need to test the 4th version of Firelis.

    Internet Explorer 8

    image
    Unfortunately, the browser from Microsoft continues to corrupt in this test. Only 345 out of 574 subtests have been completed. Browser users will not be able to enjoy the many features of CSS3. The only consolation is that not all specifications are recommended by W3C. We’ll have to watch how things are going with the test versions of the new Explorer.

    Internet Explorer 9 PP4

    image
    The test version of Internet Explorer 9 allows web designers to be optimistic about the future, realizing that soon they will not have to deal with optimizing pages for each browser. What am I talking about? Oh, for sure. In general, Internet Explorer 9 passes the test completely without a single error.

    Safari 5.0.1

    image
    Apple's browser does a great job with the test. 574 subtests passed.

    Maxthon 3.0.15.300 RC


    The "exotic" browser is not stewed in front of eminent competitors. Test passed.

    HTML 5

    The new version of the markup language has already been around for about a year, giving rise to a lot of hopes that soon it will be possible to get rid of all kinds of "layers" such as Adobe Flash, MS Silverlight, etc. The prospect is quite tempting, given that not all devices (for example, mobile) can work with heavy plug-ins, and some OSs (for example, based on Linux) do not work correctly with them, because either the developer is not very eager to search for errors in the plug-ins (Adobe Flash), or generally does not officially support the plugin (MS Silverlight). Thus, HTML 5 can solve such problems by providing a new cross-platform standard. While the language is still being developed, but there are already some projects running on HTML 5, ranging from the beta version of YouTube video hosting, to some browser games.Niels Leenheer ( Official Website ) has developed a special test that can be run by clicking on the link .
    What is most expected of HTML 5? Elements for video , audio , canvas and geolocation . The rest of the specifications are interesting, but not so popular among the masses ( WebGL , Workers and others).
    So, let's see what each browser will show (bonus points will not be included in the result, since they have no direct relation to html5):

    Opera 10.60.3445

    image
    Opera 10.60 scores 159 points. The most popular elements such as video, audio, geolocation, canvas are supported.

    Opera 10.70.3468

    image
    The raw alpha version of Opera does not show dynamics, although, according to developers, some improvements in working with HTML 5 have been made.

    Chrome 5.0.375.125 Stable

    image
    The brainchild of Google shows very good results, gaining 197 points. All elements most expected by most users are supported.

    Google Chrome 6.0.484 Canary

    image
    Finally, the “canary” appeared in our testing, having already scored 217 points out of 300. Google and the Chromium developers are not asleep, but continue to add new features.
    It should be noted that by prescribing certain parameters, you can also increase the result of Google Chrome by adding WebGL support to the browser. By default, due to "dampness" the feature specification is disabled.

    Firefox 3.6.8

    image
    Mozilla Firefox supports all the basic "goodies" of the new markup language, gaining 139 points, which is good news.

    Firefox 4.0 b2

    image
    The second beta version of Firefox 4 demonstrates what work the developers of the browser did. Firefox 4 is gaining 189 points.
    It should be noted that by prescribing certain parameters, you can also increase the result of Mozilla Firefox by adding WebGL support to the browser. By default, due to "dampness" the feature specification is disabled.

    Internet Explorer 8

    image
    A fresh stable version of Internet Explorer shows that it is absolutely not ready to work with the new version of the HTML markup language. The browser supports only some HTML 5 specifications, while not supporting the most popular functions, such as video, audio, canvas, geolocation.

    Internet Explorer 9 PP4

    image
    In the alpha version of Microsoft’s new browser, certain shifts are visible in the implementation of HTML 5 support. The browser supports canvas, video, audio, while it flatly refuses to work with geolocation and scores only 96 points out of 300 possible.

    Safari 5.0.1

    image
    The "Apple" browser handles the test slightly worse than Google Chrome 5, but Opera 10.60, Firefox 3.6.8, and both versions of Explorer are better. The browser supports all popular elements, gaining 165 points.
    On Mac OS X:
    image
    Thanks to the esteemed farcaller user

    Maxthon 3.0.15.300 RC

    image
    SUDDENLY, the Chinese browser managed to surpass even Safari under Windows, gaining 166 points, although there is no support for geolocation. Given this result from Maxthon, other browsers should take into account that this browser may well occupy the Chinese market. And there are more than a billion of them (Chinese) (although, naturally, far fewer than a billion Chinese have access to the Network).
    In Trident mode, Maxthon scores 27 points, which is similar to the result of Internet Explorer 8.

    Here is a table showing the browser’s readiness for modern trends:
    image
    For more details, see the specifications here

    What intermediate conclusion can be drawn from all this disgrace? The browser market has revived. Until 2008, there was some lethargy, but after the laboratories of Opera Software ASA and Apple began to report that the Presto and WebKit engines began to pass Acid3, a new phase of the browser war began to flare up. Microsoft dozed exactly until there was a serious tendency to capture the market from Google Chrome. Lazily sleeping Ognelis started up when Google looked away from Firefox and focused on its Chrome. And when support for extensions was pushed into Chrome, it became completely dangerous. Web standards began to be actively implemented in Opera, Chrome, Safari browsers. Mozilla was also in a hurry, having begun to saw through its browser, expanding support for HTML 5. Microsoft had the most work.for fall . And competing browsers in their stable branches have many elements that IE does not even contain in the test one. Of course, many can say that due to clumsy support for web standards, IE will die, but so far the dynamics are such that the share of Microsoft’s browser has only been growing for the last two months. Of course, against the background of the emergence of some resources using HTML 5 (browser games), Internet Explorer’s share will drop slightly by the fall + possible hype with the release of Google Chrome 6, which will most likely take place in late summer and early autumn. These factors will take away the share of MS IE, but it is difficult to guess anything. It is impossible to talk about the “young” test participant - Maxthon, as serious competition on a global scale, and Maxthon itself uses the WebKit rendering engine, developed primarily by Apple, Nokia and Google, and not its own. So, although this is not an upgraded “clone” of other browsers (such as Wyzo and CometBird in Firefox, or SRWare Iron and Yandex Chrome in the case of Google Chrome), there is no chance of imposing serious competition on the browser.
    In general, the browser war has subsided in the summer, shaken with the release of beta versions of Firefox. But the intermediate output is already turning into a full-fledged output, so I will interrupt my outflows with a new category of testing, namely the performance of the JavaScript engine.


    Performance


    So, recently, many Internet users have been paying attention to how quickly and efficiently browsers process Javascript. The question is, what for do we have any difference in milliseconds?
    one). Milliseconds may well turn into seconds when the computer is loaded with background processes or a large number of browser tabs / windows.
    2). With the development of HTML 5, there are opportunities for creating, for example, games that will use JavaScript. And given the global nature of the games, there can be many scripts on the page. All of them must be processed as quickly as possible, because the brakes will "eat up" the gamer. And good JS processing engines will untie the hands of developers who may not care about performance, even creating 3D games.

    Now let's look at how fast browsers are ready to process JavaScript. As a test, I refused to take “interested” tests like IE Testdrive, Mozilla Dromaeo, Google V8 Benchmark, Apple Sunspider Benchmark Test. Of course, I understand that the code for all these tests is open and everyone who knows programming can check the “twist”, but the purpose of these tests is somewhat different. Each of these tests was developed to test the engine of a particular browser and takes into account the features of a particular engine. Therefore, other browsers may show lower results. To avoid this injustice, I decided to take the “neutral” Peacekeeper Futuremark test. From the outside, I see no interest in Futuremark, so I think this is one of the most objective tests.
    Testing technique:
    I disconnect from loading many Windows services that are not needed at the time of testing (Windows Time Service, Windows Defender, print services, etc.), I also clean up the startup, leaving only the video download. That is, all sorts of IMs, antiviruses, firewalls, etc. will be disabled. Of course, in reality, such conditions are unlikely and not recommended, but I want to avoid injustice, such as during testing Opera other processes will be idle, and during testing Internet Explorer, the antivirus will start checking the computer, the update service will start to install patches, which will negatively affect on the results.
    I also clear all temporary browser files that have accumulated during previous tests, and the computer will reboot before each test, and within three to four minutes after the demonstration of the desktop I will wait for the system to load. Ideally, it would be nice to run each test several times for each browser and display the average value, but I really do not want to spend half an hour on each web-based adventure. During testing, each browser will be launched in full screen mode. All browsers will be bundled out of the box, that is, all settings will be reset to default. In general, everything is ready.

    Go!
    Opera 10.60.3445

    image
    Opera 10.60 scores 7856 points. Pretty good result. The engine performs best in rendering speed, displaying complex graphics, as well as working with text.

    Opera 10.70.3468

    image
    The alpha version of Opera shows a slight regression, gaining 7720 points.

    Chrome 5.0.375.125 Stable

    image
    In general, Google Chrome lagged far behind Opera, although in the Data subtest it showed slightly better results than the Norwegian. The result is 5735 points.

    Google Chrome 6.0.484 Canary

    image
    Based on the test, it is clear that Chrome developers were able to optimize the V8 JavaScript engine, adding about 800 points to the “stable” result, which is a lot.

    Firefox 3.6.8

    Let's see what the rather old TraceMonkey engine shows us.
    image
    As you can see from the test, the engine is not just old, it generally has a beard that has grown compared to the ultra-modern Carakan and V8 engines. We pin our hopes on JägerMonkey, which is built into version 4 of Firefox.

    Firefox 4.0 b2

    image
    Enabling part of the Safari Nitro engine code affected performance by increasing the result to 2931 parrots. But Mozilla has a lot of work, and the competitors are on the alert.

    Internet Explorer 8

    Let's look at Internet Explorer 8, which fails in all tests. Can it rehabilitate itself in a performance test?
    image
    As it turned out, no. The browser cannot render canvas graphics at all. Complete failure with 799 points.

    Internet Explorer 9 PP4

    Judging by the test results of Microsoft and Apple (SunSpider), IE 9 is just a monster in Javascript processing. Let's take a look at the test of a neutral manufacturer:
    image
    Not everything is as rosy as beautiful fish and asteroids show. Apparently, Peacekeeper does not use graphics that require a GPU connection, so the ultra-powerful Chakra engine showed fairly average results, slightly surpassing Firefox 3.6.8, which is about to retire. 2450 points. Weak for the alpha version, weak.

    Safari 5.0.1

    image
    Apple, apparently, was not particularly concerned about optimizing the Nitro engine for Windows (which is indirectly hinted at on the official website), so Safari does not pursue performance. 3875 points:

    Maxthon 3.0.15.300 RC

    image
    Hmm, Futuremark, apparently, is not familiar with the Chinese browser, so without further ado, she slyly identified the name by the WebKit rendering engine, calling it Safari unknown. I was not able to find out the name of the Javascript engine used in Maxthon, but the Chinese web-based scrambler scored 4519 points, and even managed to make all the opponents by the Data test, having gained 14119 points. It is possible that the Chinese have optimized the Google V8 engine, because it is the only one out of the Big Five that works well with the Data subtest.
    Overall classification:
    image

    Well, the results are mixed. The proprietary Carakan engine, developed by Opera Software ASA engineers, turned out to be a trump ace of the Norwegians, so the phrase that Opera is the fastest browser hangs on the site quite deservedly. Google Chrome, apparently, does not seek to “make” Opera in synthetic tests, but at the same time does not want to miss the competitor far ahead. Maxthon is a dark horse, which may well be in sudden competition with the Big Five. Internet Explorer 8 has clearly died and trump cards, in addition to some protection mechanisms, such as SmartScreen and InPrivate, does not. And in general, atas speed. IE 9 is mixed. We must not forget that the browser shows good results in the "naked" form. What will happen when they fasten the GUI with fist whistles, protection mechanisms, plugins such as web fragments, RSS aggregators, etc. P? The advantage of supporting the power of the GPU is quite impressive, but in Chromium for two months already, using the parameters, you can screw an experimental gpu plug-in, which, however, I could not get the smog due to the heat, molten brains and cruel glitches from smog. Opera has something strange with Vega graphics libraries, which are also rumored to be related to the future implementation of the graphics plugin for working with the GPU (rumor level, I repeat. I could not find any official confirmation from the experts). In general, while the final version of IE 9 rolls out, other browsers will not lag behind. Firefox, as before, is configured for user friendliness and the ability to customize the browser to fit your needs. Mozilla does not pay much attention to ultra-speed, and even a hand does not raise a rebuke. Still, many extensions for Firefox have no analogues in Chrome (which probably has the second most powerful extension database), and therefore, our beloved for openness, Ognelis will not lose the love of users. Let's hope that the developers will deal with the problems associated with changing the old interface to a new GUI, and will take up the optimization of the engine. But this is optional. IMHO, the goal of Firefox is a little different, and browser users, I hope, will agree with me ;-) Safari on Windows keeps the brand, providing convenience to Mac users running Windows. Engineers are speeding up (although not as zealously as in Google and Opera SASA), the functions are "screwing up", in general, I see no particular reason for Safari users to change their browser. As a result, all new developments show good results. I would like to thank Microsoft,

    NB The esteemed Habraiser Aux found an entry on the official blog of the Opera kernel developers about the use of Vega graphics libraries and hardware acceleration.

    NB 2 The respected GreLI Habraiser found confirmation that hardware acceleration will be enabled in future Opera builds.
    Are you going to include hardware acceleration in the next builds of Opera?

    Internal assemblies with hardware acceleration have been around for a long time, but the technology is quite thin, so the necessary fine-tuning and ensuring proper security - after all, this is iron and there are a lot of potential holes.

    Expect sooner in any of the major versions


    So the browser war is in full swing =)




    PS The article is written in a somewhat simple language, because I wanted to show it to interested friends who are not very versed in a computer device.

    PPS It’s possible that I add an ECMA-262 compliance test.

    PPPS I will be glad to hear criticism, corrections, if necessary, I will make. Errors are likely.