More advice for the full screen mode.
I list below all topic groups, which I have done according to subjects, which they handle. You can return to this topic group by using this menu and the link Table of topic groups on the top of the each page.
Table of topic groupsFront page of CSS-guideGeneral info - Advice pages > My browser recommendations > Speed of browsers and the usage of RAM (section 2/2)

Speed of browsers and the usage of RAM

Even I test primary CSS-implementations ordinary users of browsers appreciate the generic functionality of browsers, where the speed and the reliability are some of the most important features. The usage of the RAM affects to the speed. It is meaningful also for other applications.

I tested the usage of RAM with BySoft FreeRAM. I have Windows 98, 64MB RAM and 200Mhz AMD K6 processor. I tested some pages with iMatix's Xitami 2.4d2, which is some kind of offline virtual server.

BySoft, iMatix.

I free up RAM to the same level, which was at the beginning of each test approximately 30 MB (I put into memory each time the value). I used as my test pages my CSS-site as such or some modified pages. All pages had much HTML + CSS but not at all heavy embedded objects (I had not then as new browsers as I have now).

Opera browsers use normally most less RAM when opening the browser (for some reason the need of RAM changes drastic according to the size of the used browser window), opening a new window and depending of default settings also going from one page to another page. Opera browsers have automatic/user definable RAM-cache, which can turn on/off. With the RAM-cache the situation is better for Opera, but with the automatic RAM-cache Opera browser use in some situations more RAM than MS IE 6.0. If the RAM-cache has been turned off, Opera use less RAM than MS IE.

I went through ten pages with Xitami(the difference of speed might be different in the Internet and the test might not give the exactly correct impression). I compared the usage of RAM in MS IE 6.0 and Opera browser and I got the following result:

I didn't made this test with Netscape browsers. The result would not be comparative with Netscape 4.79 because it has much more simpler CSS than MS IE and Opera. All Netscape 6.x browsers use always much RAM. 64MB with Windows 98 is too little in order to run Netscape 6.2.x fast enough. Using about 30 MB RAM in order to start the browser is too much for a browser. Netscape 6.2.x needs with Windows 98 over 64 MB RAM in order to work well (Mozilla 1.1 doesn't need so much RAM).

I a static test case Opera 5.12 took a little bit more than a minute (I some test sessions was necessary to reload some pages) and Opera 6.0 took a little bit less than a minute. Opera 6.0 1 + 1 MB RAM-cache took more than 1 minute 30 seconds. 3500 Kb (documents) + 1000 Kb (images) RAM-cache values the speed was as good as with the automatic RAM-cache. The same test with MS IE 6.0 took about 1 minute 30 seconds, which mean that downloading of these pages took about 50% more time than with Opera.

Taking off the RAM-cache from Opera drop down the speed of Opera below MS IE. It seems that MS IE has fixed size RAM-cache. With some value it might be possible to find approximately the same memory usage/speed relation, which is in MS IE 6.0.

It seems that Opera use RAM-cache especially for caching CSS. RAM-cache gives faster performance, but not free! It is necessary to think, how much RAM Opera can use. In order to empty RAM-cache, it is necessary to exit from Opera. It use small amount of RAM only with RAM-cache disabled.

The speed depends however much on used structure of Web-pages and the fact if pages are static or if they are compiled with some server application. The previous mentioned test pages use a server application and pages didn't had dynamic menus. I made following additional tests on the base on my Finnish CSS-pages, which use dynamic menus (I made at least two tests with each browsers; I free up 25 MB RAM for each browser). I changed first the CSS equal for all browsers (I give results at five second accurancy).

  1. Tests without any server application (older CSS).
  2. Tests with the Xitami 2.4d2 server (older CSS).
  3. Tests with Apache 1.3.26 server with older and newer CSS. In most tests Apache run PHP in CGI-mode by using PHP.EXE. I tested with some browser Apache also so that PHP worked as a module with a DLL-file.
  4. Tests with servers with the CSS, which I have used in 27.11.2002 (I tested only some browsers). Because all browsers get a little bit different CSS results are not fully comparable but they give information, what is the effect of different CSS in each browsers. The difference between CSS for Opera and Mozilla Gecko browsers is very small. I defined Apache to run PHP as a module.
  5. I compared also how big factor slower pages, which were complied with Xitami/Apache downloaded than pages, which didn't use any server application. Results are not fully comparable, because newer CSS is more difficult than the older.

I compared following browsers:

  1. MS IE 6.0.
  2. Opera 6.05 (together 4MB RAM-cache) (I changed pages when Opera informed in the download bar all images downloaded even if the download bar was visible because the browser don't always hide it).
  3. Opera 7.0 Beta 1 (together 4MB RAM-cache).
  4. Mozilla 1.1 (Gecko/20020826) (4 MB RAM-cache, which I set to all tested Mozilla Gecko browsers).
  5. Netscape 6.2.1 (Gecko/20011128) (the browser didn't downloaded all small images of links, which are below dynamic menus; some pages should be reloaded but I didn't reload them because the browser informed that pages have been totally downloaded).
  6. Mozilla 0.9 (Gecko/20010505) (the broser correspond approximately the Netscape 6.1 browser).
Meanings fo the background colors
golden background = the best result
green baground = good achievement
red bacground = especially bad result

MS IE 6.0 Opera 6.05 Opera 7.0 Beta 1 Mozilla 1.1 Netscape 6.2.1 Mozilla 0.9
1. Tests without any server application
1 min 10 sec 1 min 15 sec 1 min 20 sec 1 min 20-40 sec 4 min 50-51 sec 2 min 55 sec - 3 min 40 sec
2. Tests with Xitami
2 min 30-40 sec 4 min 0-10 sec 1 min 45 sec - 7 min 20 sec 2 min 10-20 sec 5 min 55 sec - 6 min 25 sec 6 min 25 sec
3. Tests with Apace 1.3.26 (PHP in CGI-mode)
3 min 2 min 40 sec 2 min 45 sec 2 min - 5 min 27 sec
4. Tests with Apace 1.3.26 (PHP as a module)
2 min 25 sec not tested 1 min 40-45 sec 1 min 40-45 sec not tested not tested
5. Tests with Apace 1.3.26 with the CSS, which I have used in 27.11.2002 (PHP as a module)
2 min 50 sec 2 min 25 sec 1 min 35 sec 1 min 40-45 sec not tested not tested
6. Tests with Xitami with the CSS, which I have used in 27.11.2002
not tested not tested 2 min 15 sec - 5 min 40 sec not tested not tested not tested
Factors how much Xitami slowered the download processes between tests 1. and 2.
2,2 3,5 - 1,5 - -
Factors how much Apache slowered the download processes between test 1. and 3.
2,7 2,1 2,1 1,3 - -
Factors how much Apache slowered the download processes between test 1. and 4.
2,1 not tested 1,3 1,1-1,3 not tested not tested
Total times from test 1. and 2.
about 3 min 35 sec about 5 min 20 sec - about 3 min 45 sec about 11 min about 9 min
Total times from test 1. and 3.
about 4 min 10 sec about 3 min 55 sec about 4 min about 3 min 15 sec - about 9 min
Total times from test 1. and 4.
about 3 min 25 sec not tested about 3 min 0-5 sec about 3 min 10 sec not tested not tested


  1. Netscape 6.2.1 refused go with Apache to the start page of the tests.
  2. I couldn't get from tests which use the same CSS for all browser reasonable results with previous mentioned browsers. That's why I didn't give slowing factors. It seems that with certain CSS Netscape 6.1x-6.2x level Mozilla Gecko browsers have serious memory handling problems. With a little bit different CSS Mozilla 0.9 downloaded the test case about in 5-8 min and Netscape 6.2.1 about in 5-6 min.

  3. Opera 7.0 Beta 2 was about 10 sec slower without a server application than Opera 7.0 Beta 1. Because I have changed the structure so much other tests can't be done.

  4. In Opera 7.0 Beta 1 the way how Apache use PHP is remarkable (if it is in CGI-mode or as a module). When Apache has been configured correctly and PHP works as a module, Opera 7.0 Beta 1 is together with Mozilla 1.1 the fastest browser. If if PHP is in CGI mode, Opera is just an avarage level browser. The speed difference is 1,6 times.

  5. The problem of Opera is that Opera doesn't communicate well with like certain server applications. They slow down the rendering speed much more than any other browser. Opera 6.05 rendered guide pages of my CSS-pages with Xitami about two time slower than Mozilla 1.1 and MS IE 6.0 for Windows and slower than any tested Mozilla Gecko browser (Mozilla 1.1 suffers least server application). I tested Opera 7.0 Beta 1 with Xitami in another hard disk. That browser didn't gave consistent result. Sometimes it downloaded pages very fast, sometimes about as fast as MS IE and sometimes slow. I tested with two CSS. With the newer Opera had less problems. Also Opera 7.0 Beta 1 has some difficulties with Xitami. When I tested with Beta 2 it stopped just few times. In general Beta 2 had problems only if I the cache was empty. The problem is almost fixed in that version. Below is an e-mail answer from Opera Sofware, which concerns Opera and Xitami:

    ...Xitami is unable to handle a core HTTP 1.1 feature which Opera implements fully: Pipelining, that is, sending multiple requests to the server over a single connection, without waiting for the result of the previous request. Even sending 3 requests immediately after each other on the same connection, only returns 2 responses, the third never arrives, and Opera have to send it again after the connection times out (after 30 seconds on xitami.com)

    There is not much we can do about this, except possibly finetune our non-compliance heuristics, or just classify all Xitami servers as HTTP 1.1 non-compliant, as far as pipelining is concerned...

    ...Pipelining is a sore point for many servers.

    According to an e-mail Opera software has programmed around a bug in the Xitami server like it has done for bugs in few other servers. The problem with Xitami will be fixed in Opera 7.0. In Opera 7.0 Beta 2 the functionality has almost been fixed. Even if Opera browsers have problems with Xitami (fortunately Xitami is not very common used server), they are however in many situations quite fast. In one matter Opera 5.x+ is the fastest browser in my Web-site. It scrolls my dynamic menus much faster than than any other modern browser. On the base on offline tests Opera 7.x should download my pages fastest also in the Internet and it should be in all aspects the fastest browser in my Web-pages.

Because the speed is depending on so many factors, which browser is the fastest in average Web-page is impossible to rank. Because Apache is the most used server and PHP and other extensions works in it in most cases as modules, I presume that Opera 7.0 Beta 1 and Mozilla 1.1 download in most cases Web-pages fastest. If Opera Software gets fixed some bugs in Opera 7.0 Beta 1, Opera 7.0 has possibilities to be the fastest (full-featured) browser on the earth without any exceptions.

I tested with Opera 6.01 how the browser frees up RAM after closing the application. Opera 6.01 leaves less "trash" in the RAM than MS IE 6.0 for Windows.

Speaking about Opera as a fast browser for low resources is questionable, because Opera browsers use sometimes very little RAM and sometimes quite much - and it doesn't download Web-pages always fast.

Also loading of images affects to the speed of browsers. Opera has a button to prevent downloading images/download all images/ download only such images, which are in the cache. The button is useful especially situations, where the after the visiting the start page the user of Opera sets to download also those images, which are in the cache. If the user can use this button optimal, especial with modem connections the visitors can save much time - and money!

Mozilla 1.1/Netscape 7.0 starts without the quick launch slower than Opera 6.x and MS IE 6.0 for Windows, but not extremely slow like Netscape 6.x browsers. Netscape 7.0 doesn't anymore download the annoying "Activation" (like all Netscape 6.x series browsers do), when it starts approximately at the same time as Mozilla 1.1. Mozilla 1.1 and Netscape 7.0 can be installed with the quick launch option. Because MS IE is a part of the operating system it has also in a way quick launch. Starting Netscape/Mozilla with quick launch can that's way compare with starting times of MS IE. With the quick launch new Netscape/Mozilla starts faster than MS IE and Opera. But reasonable usage of it require in the computer more RAM than 64MB.

When I went to the Internet Netscape went to the site of AOL/Netscape and asked to register into some services. I hate this kind of commercial feature. Users of Mozilla can be free of this "service".

Netscape 6.x browsers belong to the trash folder - all of them start slow, some of them work quite slow and all of them use too much RAM compared to the speed, which they offer. Mozilla 1.1/ Netscape 7.0 are quite fast browsers and they use reasonable way RAM (I don't have tested the usage of RAM but it works in computers, which have 64 RAM). Netscape 4.79 use RAM quite reasonable and it might be reasonable browser in some special situations.

Copyright Tapio Markula 1999-2003, Salo, Finland
(table of topic groups, E-mail - add Tapio.Markula)
( tapio.markula@nic.fi)
Not into the public use without my permission. If you want to use my pages at some way read the copyrights.
Get Expression!

An editor for creating standard-compliant HTML and XML documents. Documents of this site have in most cases checked with HTML Tidy by Dave Ragget (W3C) and randomly with the official MarkUp Validation Service of W3C. The syntax of most pages in this site should conform with the XHTML 1.0 specification of W3C. Test this page!
Information about browsers, which render or print this site best.
[Get Opera!] [Get Mozilla!]
CSS-guide has been last edited 07.08.2004