journal: win

Semi-Case Study: Which Windows browser is the biggest memory hog?

By now, most of my readers (all 5 of you) have heard that Firefox leaks memory, at least on Windows. I can attest to this, as I have seen it hit 6 figures in Task Manager almost everytime it’s been open. I also saw that Internet Explorer could take up a similar amount of memory, but also seemed to take up the middle range of the 5-digit count. I hadn’t yet seen what memory usage was like for Opera, so I decided to conduct this not-scientifical-so-you-have-no-grounds-to-do-anything-legal-wise-to-us study. This study was in no way as thorough as it could have been, but tuff cookies.

The bare essentials

The system that was used to conduct this study is as follows:

  • 3.0 GHz Pentium 4 530 (Northwood Core; 1 MB L2 Cache and HyperThreading)
  • 2.5 GB of RAM (PC3200; DDR400; 2x 1 GB and 1x 512 MB)
  • ATi Radeon X850 (256 MB; AGP)
  • Monitor 1: 17” CRT (1280 by 1024; 60 Hz); Monitor 2: 15” LCD (1024 by 768; 75 Hz though I’m told that’s irrelevant)

Quick Facts:
All measurements were taken using Task Manager. These measurements were taken before any of the browsers were launched.

  • Total Memory in Kilobytes: 2620908
  • Available Memory in Kilobytes: 21426161
  • Other running applications: Notepad

1 Memory varied by a few hundred kilobytes

The about boxes for each browser can be seen below:

About Firefox
About Internet Explorer

About Opera

The Procedure

Each browser will be tested one at a time.

  1. Open browser
  2. take memory usage amount from task manager (will be open throughout)
  3. Open three tabs
  4. Navigate First tab to http://www.dtgeeks.com
  5. Navigate Second tab to en.wikipedia.org and search for the article WWII
  6. Navigate Third tab to http://www.dontclick.it
  7. Take memory usage
  8. Close all tabs but one and click Home
  9. Take memory usage
  10. close browser

The Results

One thing to keep in mind with these results is that Opera and Internet Explorer are both beta applications, whilst Firefox is a stable build. I did this just because I am too lazy to figure out the whole “nightly” system and get a comparable version of Firefox to the other two.

Special thanks to Pilky, who made the graphs up for me because I lack CS2 smile

Fresh Run

Uploaded Image

This test is to see how much RAM is taken up by a fresh browser session opened to a blank page. To do this, I set each browser’s home page to about:blank and closed each browser, making sure the process for each browser closed. I then opened the browser, waited for all loading to finish, and took memory readings.

In Use

Uploaded Image

For this test, I opened up two additional tabs in the same browser window and session as I used for the first test and navigated the first, second, and third tab to http://www.dtgeeks.com, en.wikipedia.org(I then searched for “WWII"), and dontclick.it respectfully. I then took memory readings.

Why those sites?

I selected http://www.dtgeeks.com because I feel it represents the average web page: medium CSS usage, XHTML, etc.

I selected the English Wikipedia article on WWII because I feel it represents a web page that is mostly text with just a few images; it is representative of what someone conducting research may come across often.

I selected dontclick.it because I feel it is a good representative of an entirely flash-based web page, which is what many corporate web sites use.

Closing Up

Uploaded Image

This test was to see how well each browser would release memory it wasn’t using. I closed all tabs and hit “Home” which took me back to about:blank. I then took memory readings.

Conclusion

As you can see, Firefox is the winner, hands down. This is to be expected, though, as both Internet Explorer 7 and Opera 9 are betas, while Firefox is a stable release. This also doesn’t take long-duration use, a test that could very well expose the Firefox Memory Leak. I will do a much more thorough test once both Internet Explorer and Opera are out of beta. Until then, I’ll be working on a series of previews of Microsoft Office 2007 Beta 2.


« Previous · win journal · Next »

thinkback

1.

I’m kinda surprised at the result.  I thought my Mozilla was a bit of a memory hog compared to IE (especially after leaving them running overnight).

2.

Hey, cool idea for an article!

Safari on OS X 10.4.6
G5 Quad with 4.5 gigs RAM

Start-up with blank page = 31 megs
In use = 46 megs
Closing up= 46 megs

Safari doesn’t seem to release any RAM until I quit the application. I’m guessing it sees how much RAM I have and just uses as much cache as it wants until another application starts wanting that RAM.

Safari has WebKit and IE for Windows has the same thing, right?

So isn’t there another aspect of those two browsers that is always in RAM and isn’t represented by the task manager or activity monitor in OS X?

3.

BTW, nice charts! smile

4.

Mac Fan, if an application uses a particular system API such as WebKit or AppKit, it still has to load all the requisite code into its own memory space.  Yes, there are system calls it can make that its authors don’t have to write before-hand, but that doesn’t mean there’s always some instance of WebKit running somewhere.

5.

A note about this test. Opera 9 tested in this article was the tech preview (aka alpha) 2 release

System specs:
athlon xp 2000
512 meg ram
64 meg pny geforce 4mx 440

My stats for opera 9, build 8414
16,540k startup
42,412k used
40,196k all tabs closed and about:blank

ie7 beta 2 refresh (5346.5)
27,112k startup
58,100k used
42,796k all tabs closed and about:blank

Firefox 1.5.0.3 (this was both tweaked to cut down on ram usage, and has 9 extensions running.)

24,072 startup
38,248 used
32,488 all tabs closed and about:blank

6.

A great test, and I’m suprised too. Did your firefox have any extensions installed? I know that whenever I get mad and want it to run faster, it seems that removing them does the trick.

7.

The firefox used in the test did not have any extensions. It was performed on the Guest account on my computer, which is as close to a vanilla install of XP (and my apps) as I can get it.

Page 1 of 1 pages

respond

Have an account? Log in to leave your comments!

Commenting is not available in this weblog entry.