Firefox 3 May Be More Memory Efficient Than Either IE or Opera

Edy52285 writes "Ars Technica has an article showing benchmarks pitting Firefox 3 Beta 4 against other browsers. Contenders include IE7, Firefox 2, Opera 9.5 Beta, and Safari 3.0.4 Beta. The piece includes a graph depicting FF3's memory usage well below that of the other browsers. The in-testing browser even trumps Opera, which has long been regarded as the fastest browser around."

I knew IE7 was bad, but...

[TripMaster Monkey]

It's one thing to know that IE7 is a resource hog, but another thing entirely to view the graph in the article and be confronted with hard evidence of just how abysmal it is.

I'm going to print out that graph and put it on my wall. Then, when my users come to me and ask why our enterprise isn't rolling out IE7 on our systems, I can just point to it.

Based on my experience with FF2

[Chrisq]

Based on my experience with firefox 2 I would say that once you have a few plugins (cough: *adblock*) the graph will not be flat but will slowly increase. Not that this is the fault of the browser writers, but it will be many people's real world experience.

Re:Based on my experience with FF2

[pohl]

Rejoice: FF3 has some garbage collection improvements [mozilla.org] that should fix many leaks caused by browser add-ons.

Re:Graph shape

[savala]

Out of curiosity, what's the dropoff and flatline near the end of both Firefox lines on the graph? Anyone know?

From the original blog post [pavlov.net]:

For the results below we loaded 29 different web pages through 30 windows over 11 cycles (319 total page loads), always opening a new window for each page load (closing the oldest window alive once we hit 30 windows). At the end we close all the windows but one and let the browser sit for a few minutes so see if they will reclaim memory, clear short-term caches, etc.

So that is all the memory being reclaimed upon closing all but one of the windows, and then doing nothing whatsoever.

Re:Remember when people coded for small memory use

[Nibbler999]

That graph is based on 30 open windows at a time, not 'basic web browsing'.

Re:A trend is emerging...

[BarryJacobsen]

if you're still on beta two, try beta four - it's noticably faster!

Re:From the ars discussion...

[bconway]

Which graph are you looking at? On the one linked, IE has double the memory footprint of Firefox when 30 tabs are open, and doesn't reclaim any memory when they're closed.

Re:That's when testing with their own tool

[FrankNFurter]

How about the SunSpider JavaScript Benchmark [webkit.org], produced by the WebKit developers?

The latest Firefox 3 nightly beat Safari 3.1 as well as the latest WebKit nightly on my iMac (2.0 GHz C2D, 2 GB RAM). You might want to run your own tests; you'll find that Firefox 3 is pretty damn quick.

Re:A Blessing!

[PrescriptionWarning]

try AbiWord (http://www.abisource.com/download/index.phtml), there's a windows version.

course if you switch over to something like Ubuntu it would be even better, though I'd imagine that would be pretty tough to do at least until XP stops getting supported some year

Re:Graph shape

[Anonymous Coward]

You can find benchmarking results for a somewhat different test with Firefox 2 [mozillazine.org] and Firefox 3 [mozillazine.org] using tabs instead of windows. They still show the same end result: Firefox uses less memory than other browsers.

Re:Nice to know

[yamiyasha]

FF3 was in the latest Alpha Build of Ubuntu 8.04

As long as it doesn't crash...

[Anonymous Coward]

As long as it stops crashing on me, I'd be happy. Firefox used to be a fantastic browser. It was stable, fast, and secure. It is a still a great success story but it is far from stable nowadays. It crashes on me at least 3 or 4 times a day on my Ubuntu box. I have also heard similar complains from Mac users.

plugins

[lseltzer]

I've been using FF3 for months and it's definitely efficient with memory, but the graph doesn't reflect my own experience with IE7 and FF2. At the moment, for instance, on my XPSP2 system with both FF2 and IE7 running, probably for weeks, FF2 is using about 509MB and IE7 about 208MB.

Perhaps some of the differences here have to do with plugins? There are still a bunch that don't work with FF3.

Re:That's when testing with their own tool

[Anonymous Coward]

I don't know about competing browser developers. They don't seem interested in benchmarking the memory usage of different browsers for some reason. I wonder why? You look at another memory usage test [kejut.com] that concludes Opera uses slightly more memory than Firefox, and IE is really a memory hog, which interestingly matches the graph on Mozilla's benchmark.

Re:Scale?

[Anonymous Coward]

According to pavlov [flickr.com]: We purposefully put in a 3 second delay for all the pages so that would all take about the same amount of time in all browsers (as my post was about memory usage, not page load times (also, several browsers don't render all the pages correctly)) and didn't want to confuse anyone.

Re:Remember when people coded for small memory use

[mi]

That graph is based on 30 open windows at a time, not 'basic web browsing'.

My (very) significant other keeps 5-10 windows open with 4-12 tabs in each... No kidding...

Here is the top(1) entry of her firefox-session (running linux-firefox-2 on FreeBSD/amd64):

84676 i 1 96 0 1078M 613M select 1 524:47 4.98% firefox-bin

My own (native) session uses 2.5 times less... In other words — "common practice" is a very loose standard :)

Re:comes at a cost

[Anonymous Coward]

Can't remember where I read it, but I recently read a description of how they achieved some of this efficiency. Much of it has to do with using a different memory allocator which avoids fragmentation.

Perhaps in TFA?

From TFA: The developers also adopted FreeBSD's jemalloc allocator, which helped reduce fragmentation and improve performance.

As to your points about the tradeoff between memory usage and CPU usage: the benchmarks also show a significant increase in browser speed, so this isn't going to be an issue.

WTF does polling have to do with expunging cache data? I doubt they use polling, but even if they did, how many times a second do you plan on expunging cache data?

Re:Graph shape

[epine]

I happened to have a Fedora system, so I stuck with FF 1.5.x right up until the first day of FF 3b1. I do a lot of work in MediaWiki environments, often pounding away the whole day in FF. Somehow, I rarely manage to have less than 50 tabs open, occasionally as many as 200, in four to eight windows scattered over four desktops.

Memory usage under 1.5.x was unbelievably bad. After a week of heavy use, it would routinely plateau in the 1-1.5 GB range, at which point it would become intolerably slow and force me to restart.

I've downloaded every FF 3 beta the day of first release, and pounded on them all.

3b1 crapped out after just over 2 weeks of heavy use. 3b2 was noticeably better, but not perfect. I wasn't thrilled with 3b3. Page transitions to previously open tabs became more sluggish, back/forward browsing was slower, and they really messed up window to window tab move (didn't take the tab history along for the ride, causing me to lose some major unsaved edits while discovering this unpleasant fact, which happily is now fixed in 3b4).

3b4 has been tremendously solid over the relatively short period since its release. Virtual 540MB, resident 330MB. That's spectacularly low by the standards of previous releases for the intensity of my use. Back/forward page transitions on aged tabs remains slower than for 3b1, but not annoyingly so. Overall, it just feels solid now.

I'm having trouble comprehending that *anyone* once said Firefox had no serious memory leaks. Say what? Firefox 1.5 was the Ginny Sacramoni of web browsers. I'm happy to confirm that Firefox has successfully excised the 90-pound mole from its waddling derriere.

Re:That's when testing with their own tool

[Constantine XVI]

On Windows, Athlon XP 2400+ (2GHz), 1GB Ram, Firefox3 beta 4 vs WebKit nightly 31109 (today)
FF3b4:http://preview.tinyurl.com/2xwkm3 [tinyurl.com] 7001.8 ms
WebKit:http://preview.tinyurl.com/2cjjfc [tinyurl.com] 8503.4 ms

Re:Crash

[CastrTroy]

Firefox 2 lets you reopen closed tabs, so I imagine that Firefox 3 would also have that functionality.

JavaScript performance

[Niten]

Reduced memory usage is great, but if you're more interested in speed you should take a look at Firefox 3b4's results on the Sunspider JavaScript benchmark, where testers commonly found that it performed twice as well as the latest Opera beta, and nearly three times as fast as Firefox 2 [mozillalinks.org].

I haven't yet heard anything definitive about Gecko's performance in FF3 with respect to FF2 or the rendering engines in other major web browsers, but from my own experience with the betas I can subjectively say "it's fast"; if I'm missing out on speed using FF3b4 instead of the latest WebKit, I can't tell the difference myself.

And Beta 4 is quite stable, to boot. Mozilla really pulled out all the stops on this one... unless you have incompatible extensions holding you back, do yourself a favor and upgrade now.

Re:A Blessing!

[sznupi]

With 1GHz I'd still choose Opera - definatelly feels _much_ more snappy with many tabs open.

Re:Remember when people coded for small memory use

[justthinkit]

Ok, I normally use Opera and have avoided Firefox due to the community voice regarding memory use. So I decided to give FF2 & FF3beta4 a try today. Here is my usage stats:

Browsers: Firefox v2.0.0.12 (no plugins), IE v6.0.2900.2180 (I can't stand the look of IEv7), Opera v9.23, Firefox v3Beta4. Caches cleared before test.

Memory used on initial load:
FF2 IE O FF3
26 17 45 27

Memory used after loading each of these in a new tab (window for IEv6):
FF2 IE O FF3
28 25 58 34 http://www.firefox.com/
45 46 76 52 http://us.imdb.com/
68 71 86 74* http://www.espn.com/
73 80 89 73 http://www.pcmag.com/
76 82 91 75 http://www.extremetech.com/
79 86 95 79 http://www.wired.com/
88 98 104 87 http://www.cnn.com/
97 116 108 93 http://www.amazon.com/
99 124 108 95 http://www.slashdot.org/
104 148 111 101 http://www.google.com/ig
* - 74, dropping to 67 after 10 seconds

After closing all tabs:
FF2 IE O FF3
66 67 104 58

Amount released when program is shutdown (as shown in Task Manager):
FF2 IE O FF3
56 53 100 47

Amount not released (as per TM):
FF2 IE O FF3
10 14 4 11

Note: Browsers (espec. IE) don't necessarily show all memory used by their entry in Task Manager so I prefer to know what memory was free before they loaded, and just as importantly after the browser in question is closed.

Comments: Ok, I was surprised how well FF2 & FF3 did in these tests. I also noticed Firefox properly rendering that slideshow-like flash thingy on espn.com (where my Opera setup doesn't show it at all). And that Opera acted pigishly :-{ I think it is time to give Firefox another trial.

NoScript makes a major impact on Firefox memory

[tjwhaynes]

I'm having trouble comprehending that *anyone* once said Firefox had no serious memory leaks. Say what? Firefox 1.5 was the Ginny Sacramoni of web browsers. I'm happy to confirm that Firefox has successfully excised the 90-pound mole from its waddling derriere.

If you ran NoScript on Firefox, you probably were entirely happy with the memory usage. Much of the memory fragmentation and leaks due to circular references was caused by Javascript, either on pages loaded or other extensions running. NoScript radically reduces the amount of Javascript being executed by your browser and therefore radically reduces the amount of memory used/fragmented/leaked.

Plus of course, the performance of page loading also improves because your browser isn't trying to execute some moronic scripts designed to track your movements and display "punch the monkey" ads.

Cheers,
Toby Haynes

Works great on OLPC

[dovgr]

I started using Opera on the OLPC as Opera is touted as the minimum resource GUI browser. I once tested FireFox 3 Beta 3 without too much expectation, and was positively surprised that it gave a feeling of quicker response than Opera. Same with FireFox Beta 4. There are still some scrolling issues and redrawing that is irritatingly slow. E.g. the mailbox overview frame of gmail.

Re:That chart is odd...

[whitehatlurker]

You're right, it's anecdotal. Next time you minimise IE, check its VM size, not Mem usage.

The chart was generated by running the same test, which may or may not measure your browsing habits, on all browsers and seeing how they reacted.

As an Opera user, I am surprised, but hope that the release version of Opera 9.5x will be better than the beta with respect to this. The other thing is FF 3.0's Javascript speed, which has improved remarkably.

Re:I knew IE7 was bad, but...

[not flu]

I have never had 50 tabs open at once. I think my limit has been around 20, but I usually do not average more than 5. 50, for real? Does not sound like a real world test to me.

200+ tabs in all windows combined is nothing unseen for me. I hate interrupting the flow of reading a page that has tons of links for example, so I open them all in new tabs (or windows) and check them out afterwards. Shoot, a gallery of images, waiting for each pic to load is going to take a couple of minutes total! Open them all up in new tabs, faster to switch between tabs than to wait for each of them to load in front of my eyes. 50 tabs is "light" usage to many users, such as myself.

Re:Why Safari 3.0.4 beta?

[Bigjeff5]

Perhaps you should RTFA?

Safari 3 and Internet Explorer 8 could not be benchmarked because they crashed during the test.

They tried using Safari 3 (non beta), it crashed so it wasn't included. It looks like the beta even crashed after a short while, but there is enough for a benchmark. It seems the most stable browser for multiple windows, Safari is not.

They tried using IE8 beta also, it crashed so it wasn't included.

Re:Remember when people coded for small memory use

[mi]

My point was not against the killing — it was against kill -9 . Regular kill is just as effective in most cases, but gives the process a chance to clean-up — inside a signal-handler [wikipedia.org]. Using -9 gives no such chances — the process never knows, what hit it. This is the common source of left-over temporary files, of orphaned shared-memory segments and other ill-effects...

Only if a process refuses to die for seconds after a regular kill, is trying the -9 justified...

Re:Crash

[CastrTroy]

Just right click on the tab bar (not on a tab), and click "Undo Close Tab".

Re:Crash

[matt_hs]

Really? I can't find any clue that it's possible, much less how to do it. Is it documented somewhere, or is it an "Easter egg" that you just have to stumble across accidentally (or learn from someone else)?

Presuming you're not joking, look under History to Recently Closed Tabs.
Firefox 2.0.0.12. No special plug-ins, add-ons, etc. etc. etc.

Re:plugins

[AKAImBatman]

Those system DLLs are shared by multiple programs (at least in theory) and thus have their memory usage accounted for separate from the programs that use them. In fact, the memory counts against the operating system. Microsoft uses this fact to hide much of IE's memory in DLLs that have been installed as part of the OS.

So unless you have tools to pick apart where your OS's memory is going, you're going to get bad results for IE.

Try using something like Process Explorer [microsoft.com] instead. It will give you a much better view into what memory is being used and where.

Re:I knew IE7 was bad, but...

[Anonymous Coward]

I hate interrupting the flow of reading a page that has tons of links for example, so I open them all in new tabs (or windows) and check them out afterwards.

That's the best explanation I've seen of why I (and I assume many others) have so many tabs open at once. Initially, I was surprised your comment wasn't modded up.

200+ tabs in all windows combined is nothing unseen for me.

50 tabs is "light" usage to many users, such as myself.

Holy crap! 50 tabs is close to my personal high. Does one page/article have this many links that you want to check out afterwards? If not, then maybe you need to stop procrastinating and finish reading the oldest page/article (and its links) before opening another 50 tabs. Or perhaps you should save a collection of related tabs as a "session" (Opera's term) and close them for reading later.

Re:Crash

[matt_hs]

Not joking at all. I've seen that "Recently Closed Tabs" entry in the History menu, but it's always greyed out and unusable. I just tested it by opening a new tab, selecting it to verify that it was a real tab, and closing it. The "Recently Closed Tabs" menu entry is still greyed out, although I just closed a tab. I've also had a number of other tabs open during the day, and closed them, and that "Recently Closed Tabs" thingy is always greyed out when I check it.

So how does one enable it?

(This is on a Mac Powerbook with OSX 10.4.11, if that matters. I've also seen that menu item with FF on my linux box and my wife's NT and Vista systems, and it was also greyed out there. So I'm baffled. What good is it if it can't be used? ;-)

You don't have the Estonian language pack installed, do you??? :-) https://bugs.launchpad.net/ubuntu/+source/firefox/+bug/129749 [launchpad.net]Ubuntu Bug 129749 discusses the issue (although I understand yours is on OSX . . .)

There are a few bug reports I found whilst Googling and also looking in Google Groups. Some IceWeasel Bug ID #400704 commentary points to not having a home page defined; one user said defining the home page to be "about:blank" fixed it. More promisingly (I think) is that under about:config, there is an entry called browser.sessionstore.enabled. Try checking it and turn it on if it's off. http://groups.google.com/group/mozilla.support.firefox/browse_thread/thread/4b9ba0eb24229c34/d4a1b0188a9e17ac?hl=en&lnk=st&q=firefox+%22recently+closed+tabs%22+(%22grayed%22+OR+%22greyed%22)#d4a1b0188a9e17ac [google.com]

Just a guess . . . since I haven't experienced it myself.

Re:I'm certainly impressed

[Kintanon]

To remove the home button right click on Home, click Customize, then click on the Home button and Drag it into the big box of icon things that opens up.
That will pull it off of the toolbar.

Re:Crash

[jc42]

You don't have the Estonian language pack installed, do you??? :-)

No, but I do have Finnish Extended and Swedish Pro, which are pretty similar. (Some linguists argue that Estonian is a dialect of Finnish, but the Finns insist it isn't because Estonian is incomprehensible to them. ;-) I also have Russian, Hebrew, Arabic, Greek (Polytonic), and 3 Chinese packs. I wonder if this might cause problems? I know that I'm constantly stumbling across inexplicable, spontaneous switches of language. This is especially annoying when it switches to Swedish or Finnish, because they're nearly the same as U.S. Extended, and it sometimes takes me a while to realize why things aren't working right.

Some IceWeasel Bug ID #400704 commentary points to not having a home page defined; ...

It's defined here, as http://en.wikipedia.org/wiki/Special:Random [wikipedia.org], which is one of my favorite "pages". ... under about:config, there is an entry called browser.sessionstore.enabled. Try checking it and turn it on if it's off.

It's there, and it was on.

Maybe I'll try experimenting some more. I did ask google, of course, and while it finds lots of pages that mention undoing a tab delete in firefox, the first couple dozen don't seem to mention how they do it. They just say how useful it is, which I'd agree with, because I'm always closing the wrong tab. It probably has a lot to do with having a dozen browsers installed (for web testing purposes), and no two of them handle tabs quite the same.

Re:Crash

[Anonymous Coward]

I don't know what's happening with your install, but I suspect you're misinterpreting what's happening. I'll address a few things.

The tab bar disappears by default when only one tab is open. This is for those who care about maximizing screen real estate by eliminating unnecessary components. If you would like to disable this, you can go to Tools > Options or Edit > Preferences (I don't know where it's found on mac builds) to get to the preferences dialog. Go to the "Tabs" section of the preferences and check "Always show tab bar."

As for the "undo close tabs" feature [mozilla.com], it is documented in the Firefox help documentation. Go to "Closing and Restoring Tabs" under "Tabbed Browsing" in the help viewer. You can find additions to the featureset in the release notes of each release.

As for your experience with this feature, try this: Open a new window. Assuming you only have one home page, there should be one tab with your home page loaded. Now open another tab and navigate to some website. Click on one of the links on that page. Open two new tabs. Leave the first one blank and navigate to some website in the second new tab. You should have four tabs open now: your home page, one web page that contains one entry in its history that allows you to navigate back to the first website of your choice, one blank tab, and another tab at another website of your choice. Now, in this order, close tabs 4, 3, and 2.

If you go to History > Recently Closed Tabs, there should be two entries, each containing the text for the titles of the two web pages that were present in the tabs you closed at the time that you closed them. Notice that the blank tab is not present. Firefox discards blank tabs with no history. If you bring up the context menu on the tab bar and select "Undo Close Tab," it will bring up the most recently closed tab. It should be the web page you navigated to from the first website you visited, with the history intact (i.e., you should be able to click "Back" and it will navigate back to that first website you entered). If you repeat this process, it will reopen the tab that you first closed. Note how Firefox opens the tabs in reverse of the order they were closed, that is, the most recently closed tabs are reopened first (last-in, first-out [wikipedia.org] order).

By default, Firefox limits the tab stack to 10 entries, although this can be changed by editing a config entry.

HTH HAND