Slashdot Mirror
Real-World Firefox 3 Memory Usage Leads the Field
An anonymous reader writes "The author developed a program to snapshot memory usage per process every 3 seconds on Windows. Using this he recorded 3 hours of memory usage for five different browsers under real-world usage scenarios: Safari 3.1, Firefox 3, Flock 1.2 (a browser based on Firefox 2), Opera 9.5, and Internet Explorer 8. A million data points indicate that Firefox 3 has a surprising advantage over the other browsers tested. These are real-world tests and not contrived benchmarks."
The article states FF3 is an improvement over FF2, without offering data points for FF2. However, it also mentions Flock is based on FF2, so I'm guessing they've assumed the Flock data is representative for FF2.
Am I the only one that thinks although firefox 3 is much faster, firefox 2 was much more stable? I'm running it on vista, XP sp3, and ubuntu machines and FF3 crashes on all 3 of them.
Don't get overexcited just yet. Let me quote some of the most important parts of the article that were completely overlooked in the summary for some reason:
"These results are from opening Memory Watcher and then using the browser between 9,000 and 11,000 seconds (close to 3 hours). Each browser is tested in a separate session, and there are brief periods of inactivity throughout the time period. [...] The above profiles are not a direct comparison in any way, but they offer a visualization of trending in the memory behavior of the layout engines and interfaces. [...]These aren't stress tests, and I probably never went over 4 windows in each browser, with at most 3 tabs in each window. [...] An automation script will never give the same insight into performance over time as will this sort of profile." [emphasis added]
In other words, it is evident that there was no guarantee whatsoever that every browser would display exactly the same sequence of web pages. It is easy to jump to conclusions that if Firefox has used the least memory then it must "[have] a surprising advantage over the other browsers." But is it a logical course of reasoning? Or only a post hoc ergo propter hoc fallacy combined with wishful thinking? The truth is that the amount of memory used during an hour of downloading web pages is strongly correlated with the speed of downloading and displaying said web pages. Is it the case that Firefox couldn't download, format and display pages as quickly as Internet Explorer because of the native Windows internal API hooks that help Explorer work faster than any independent browser could possible aspire to? That is quite possible. Unfortunately the results of that experiment are inconclusive and the methodology was unreliable.
Karma: Positive (probably because of superiour intellect)
Final memory usage in MB
Safari 636.9
Firefox 3 111.8
Flock (Firefox 2) 191.9
Opera 9.5 190.6
Internet Explorer 194.4
I can't be arsed to do this properly. There are a lot of graphs and stuff. But here are the text bits. Pure copy and paste. No formatting. Live with it.
Web browser performance is an often talked-about and flaunted thing, but many claims are not really backed up by solid evidence. I wrote software that collected millions of data points over 14 hours of actual browsing time, and this article reveals my findings.
Problem
Many people load hundreds of web pages, sometimes at the same time, often over periods of 3+ hours. Users complain about the memory usage of Firefox, Safari, or Internet Explorer, and we need a way to identify which browsers are better at managing memory than others. Traditional benchmarks do not look at all the things you might do with a program, and we need real-world numbers over a period of hours.
Solution
I developed a Windows Forms application in .NET called Memory Watcher that "watches" the system memory numbers. It uses a timer to poll the processes every 3 seconds. It then records every number and also prints them out in a grid on the screen. This allows us to keep track of each program's memory usage over time and with real-world usage.
Memory Profiles
These results are from opening Memory Watcher and then using the browser between 9,000 and 11,000 seconds (close to 3 hours). Each browser is tested in a separate session, and there are brief periods of inactivity throughout the time period. The vertical axis is the memory used in MB, and the horizontal axis contains the memory "checkpoints" my program took (one every 3 seconds).
(Graphs and more graphs)
Benchmark Details
The above profiles are not a direct comparison in any way, but they offer a visualization of trending in the memory behavior of the layout engines and interfaces. This is not a diagnosis or bug report. Let me show some important metrics of the above results.
Browser name Exact version Time active (s)
Hours Comments
Safari 3.1.2 10,470 s
2.91 hours Normal browsing
Firefox 3.0 9,681 s
2.69 hours Normal browsing
No extensions
Flock 1.2.2 10,146 s
2.82 hours Flock is based on Firefox 2.0
No extensions other than the default
Opera 9.5 9,855 s
2.74 hours No extensions
Only browser was used
IE 8.0 10,236 s
2.84 hours Used 7.0 rendering mode
No extensions
The system is Windows Vista SP1, and the computer has 3.0+ GB of RAM. No plugins are disabled, but the Acrobat Reader and Java plugins were (presumably) not used. Flock is based on Firefox 2.0 but its memory usage is probably worse because it uses built-in extensions.
* Just regular stuff
* These aren't stress tests, and I probably never went over 4 windows in each browser, with at most 3 tabs in each window. I didn't look at many pages that are extremely heavy on images, and no "browser benchmark" style pages. Gmail was used on each browser.
* Not just pages
It is hard for a regular benchmark to "simulate" a user actually clicking on things. Interactions with the user can greatly influence memory or performance. Having a responsive browser is probably more important than just having a "fast" one at showing pages.
* Plugins included
My profiles include Flash and possibly other plugins. A browser might have memory issues with a plugin and that could cause a significant problem with the user experience. (Most Windows Vista crashes have been due to graphics cards, not Vista itself, for example.)
* Real-life usage
An automation script will never give the same insight into performance over time as will this sort of profile. As developers, we want to make programs that work wel
My wife and I share a computer. She uses mostly uses Firefox, I mostly use Opera. This is on a 64-bit Ubuntu Hardy.
I have noticed no difference in her memory usage since we upgraded to FF3. I used to regularly have to kill her browser every once in a while (maybe once or twice a week) because it was eating up all the RAM. Since we upgraded to FF3, I can see no difference in memory usage.
For example, right now FF is using 300MB resident, Opera is using 100MB. Flashblock is installed on both browsers. Granted, that's not a terribly good test considering we've been browsing to different sites, but I've found that those numbers are fairly stable. FF usually levels off in the 300-500MB range, and Opera in the 100-150 range.
YMMV.
It's always a long day... 86400 doesn't fit into a short.
Tests need to be repeatable, and when they are repeated they need to provide consistent results. If you can't provide consistent results on each subsequent test, you need to provide solid rational on why the results were outside your margin of error. Saying "Oh, I might have had one or two youtube videos up at the time" is not solid rational.
Scripts that visit the exact same pages, for the exact same time, do the exact same things across all browsers provide consistent, quantifiable results. Since everyone's browsing behavior is different no script will ever provide "accurate" results for real world usage. But then again, those scripts could be closer to my real world usage than this guys anecdotal test. Get it?
That's pretty much irrelevant to me, I switched back to Opera because of performance issues.
Memory != performance.
For example, when I open a new tab in opera the CPU doesn't register almost any change, when I open a tab in Firefox it goes almost to 100% (that's in Linux, with many extensions added, and BTW, I need those extensions to duplicate Opera's features)
"It is our choices, Harry, that show what we truly are, far more than our abilities." -- Prof. Dumbledore
In the case of Firefox, memory usage has ranged from 25MB to $MEMORY_AVAILABLE. Which sucks no matter how much you have.
If you check this fairly lengthy explanation of how memory usage was improved in FF3 you'll see that it is mostly attributed to reduced fragmentation and leaks, and smarting caching, just as you are advocating.
db
I am literally 3000 tokens away from the chaotic crossbow --Stephen
The OS was Vista, and the program was written in .Net to use the function PrivateMemorySize64. MSDN says it returns "the amount of memory which cannot be shared with other processes". It also says it's the same as the "Private Bytes" value in taskmon. Probably it means that it's the amount of memory the process received from mallocs (or rather GlobalAllocs/LocalAllocs/HeapAllocs), and which can't be assigned to some other process.
It's worth noticing that the guy bothered with a GUI and an interactive filtering option for such a simple program. I wonder whether he ever heard of CLI, because it looks like a perfect fit for this kind of program.
By the way, why not post CoralCDN links (append .nyud.net to hostname) instead of direct links when the site in question is small and likely to be Slashdotted?
Those who would give up liberty to obtain working drivers, deserve neither liberty nor working drivers.
Some, but I'm not one of them, my Opera (9.51 on XP SP3) has been running for a few days now (almost 4), with a peak of 157MB, currently at 96MB, VM of 114MB, and an I/O of almost 12GB's... opening up every site on my SpeedDial (9 sites + this one) brought me up to 122MB, VM of 140MB...
But, I honestly don't care how much it uses, because so far it hasn't impacted (noticeably) on any other software, and always starts (launches, or maximizes) instantly, and I prefer Opera's interface. And how much memory it uses isn't enough to make me prefer one browser over another... CPU usage on the other hand, might, but most of them are pretty much the same in that regard.
And after closing all those tabs, its down to 92MB, VM 110MB, peak the same.
http://dotnetperls.com.nyud.net/Content/Browser-Memory.aspx
On the off chance you're not just trolling...
Instructions are here:
http://www.kennycarlile.net/2008/06/17/disable-firefox-3-awesomebar/
Took me all of 10 seconds with google: "Firefox3 Awesomebar disable". So... I guess you'll enjoy firefox 3 with the smaller memory footprint (assuming TFM is correct).
Clones are people two.
What a fool believes, he sees, no wise man has the power to reason away.
NetworkMirror.
Ant(Dude) @ Quality Foraged Links (AQFL.net) & The Ant Farm (antfarm.ma.cx / antfarm.home.dhs.org).
Look at this article talking about memory usage in Firefox 3 by mozilla dev Parlov. There is a graph there that suggests Firefox 2 uses around 80Mbytes more memory than Firefox 3 over time.
Memory that is being shared between processes (e.g. through the use of libraries) is actually counted towards the RSS so something that is sharing a lot of pieces with another process may appear artificially high.
Based on my experience, the flock data is not representative of firefox 2. After running firefox and using tabs for a few hours, it would often be using around 300mb of my memory, while the data shows flock maxing out at 200mb.
Linux automatically uses all free memory for disk cache, and is very efficient at it. Instead of hogging memory that could be used by other process, Firefox could exclusively rely on a disk cache and rely on Linux' native and performant disk-cache to handle the in-memory caching of those files.
The best part of this is, if some other process needs the memory, Linux will simply free some memory from the cache, but the files will still be ready on the disk and the over application will still be performing well. Whereas if a 120Meg space is enforced a in-RAM cache, when memory becomes scarce, the system is at risk of paging out piece of the software (pages of code itself instead of pages holding cache) and thus make the whole system less responsive. The GC is just for JavaScript (required by design) and for DOM nodes which end up being circularly referenced (which is unavoidable). You would only need to keep the DOM nodes of the current page. Past pages are freed and don't (usually) keep DOM objects alive. Finally, 120MB is not a lot of RAM. Well, it depends. Notice that Linux is also very often used on kiosk with limited features and on old hardware which may not have huge amounts of resources.
Being able to run within a small memory space is critical for linux. Otherways, there won't be any difference with Vista.
Also a lot of problems are comming from bad Add-ons or even half-assed Browser Plugins. Flash is such a pain in the ass that can momentanily freeze the whole browser session.
Disclaimer : I run Firefox on Linux with in-memory cache disabled and using Gnash plugins instead of Flash (runs in separate process and can have autostart disabled). Adblock+ and Noscript also help avoiding that my browser loads tons of useless shit. And until recently my main desktop was a Pentium-III with 440BX chipset (a machine on which 1GiB of RAM is a rare occurence), but I didn't get any major problem even on recent distros. (Vista on the other hand had to wait)
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
I've been running Safari 3.1.1 all day (about 6 hours non-stop with 462 pages in history) and it's currently using 230.9MB RAM (fairly steady).
Opening up FF3 to check, loading 52 pages then closing down one again... 129.18MB.
Not a totally fair comparison but it's some numbers, nonetheless.
Damnit, can't use mod points in this thread now.
The way I've found to restore sane address-bar behaviour was to do the following:
Install the oldbar extension
about:config browser.urlbar.matchOnlyTyped boolean set to true.
Bingo! No more hyper-annoying address bar.
Moderation Total: -1 Troll, +3 Goat