Browser Speed Comparisons

kfrench writes "Internet browser speed tests for 'cold starts', 'warm starts', rendering CSS, rendering tables, script execution, displaying multiple images and 'history'. 'Opera seems to be the fastest browser for Windows. Firefox is not faster than Internet Explorer, except for scripting, but for standards support, security and features, it is a better choice.'"

Also

Firefox and Opera make tabs quicker than IE.

Re:Also

Well if that's you bag, how about

echo GET | netcat cnn.com 80

Whoo! Fast!

Re:Also

Maybe, but IE crashes a lot faster than Firefox ;)

Re:Also

I don't know, Opera is still the fastest I have ever seen.

The only reason I use Firefox (and I use it a lot) is that I can't split proxy servers in Opera.

The best best best part about Opera is that it doesn't check with the server when you hit the back button!!! This is the best feature in Opera, IMHO, and has saved me hundreds of hours (and that might just possibly be literal) of waiting.

When you hit 'back' in Opera, the browser simply redisplays (from the cache) what was there before. No waiting, no re-rendering, no asking 'do you want me to re-post the data from the form you filled out?' NO!!! - if I want to re-post the fucking form, I'll hit reload!

If Firefox can overcome this limitation and simply REDISPLAY the previous page, I will be a very happy man, because then I will have TWO amazing and extraordinarily handy browsers. But for now, I'll only use Firefox when I absolutely have to.

(Oh, and BTW, whoever coded the mouse gestures xpi for firefox gets a huge dollop of my undying gratitude. You made firefox usable.)

/*grabs soapbox and walks off*/

One advantage to Firefox...

...is that a motivated user can compile an optimized version [mozilla.org] or download an optimized build [moox.ws].

That option certainly isn't available in IE or Opera.

Re:One advantage to Firefox...

"That won't help on most of these tests - they test how fast the browser renders certain things, rather than how fast it downloads."

I wonder what webserver they were using to test the browsers? If using IIS...I seem to recall that IIS was 'rigged' to skip some steps normally in a browser/server conversation...and this helped IE 'look' faster that other browsers.

Dunno if this is still the case....

TCP/IP perversion

Are you thinking of Microsoft breaking TCP/IP to 'fake' faster speeds of IIS and IE? It works like this: Normally when an HTTP session ends, the TCP/IP connection is torn down by the server, which according to TCP/IP standards, involves two packets: a "disconnect request" (sent by IIS or Apache to the browser) and then a "disconnect acknowledge" (sent by the browser back to the server to acknowledge that the disconnect was received. When the client receives the "disconnect" it sends the ACK and closes up the socket on its side; when the server receives the "disconnect ACK", the connection is fully closed and the resources it uses are freed up on the server side. Under normal conditions, if the occasional ACK happens to get lost, then all that happens is that the TCP/IP socket remains open for usually about another two minutes until it times out from inactivity and gets cleaned up by the OS anyway (if you "netstat -a" you should see these hanging around for a little while).

Now, Microsoft did two things: they modified TCP/IP when in conjunction with Internet Explorer to not send the disconnect ACK, and they modified IIS to not wait until it received the ACK to close and free up the socket, but rather to close it and free up the associated resources immediately. This perversion of the very open standard on which the Internet was founded has the following effects:

• An IIS/IE exchange has fewer packets to send and less overhead when disconnecting, so artificially appears faster on stress-test benchmarks (normally a user would not feel the difference, but it makes a difference in stress-test benchmarks)
• Here's the real clincher, and this is where Microsoft's slimy brilliance shines: When an Apache server is subjected to the same stress-test of dozens or hundreds of connections per second from IE clients, because the ACK is not received, the Apache server soon ends up with hundreds of open TCP/IP sockets waiting to time out. This slows down the OS's TCP/IP handling, artificially slowing down Apache in a way that would not happen if Microsoft had used TCP/IP and not MSTCP/IP. And of course the poor Apache system is just behaving correctly according to TCP/IP.

This whole rather unethical bit of sliminess was primarily concocted to not only make IIS artificially appear faster during benchmarks, but to artificially slow Apache down (because Microsoft was getting frustrated that IIS was unable to kick the Linux/Apache servers' asses).

Re:TCP/IP perversion

You mean you don't have:

.
.
.
ExtendedStatus On
# MS hacks
DirectoryIndex home.html index.html home.htm index.htm
ACKWait MS
#ACKWait INET

?

Re:One advantage to Firefox...

RTFA, no? The Moox version of FF is in there and doesn't fair well.

How can the optimized version be WORSE...

than the "normal" version?

From TFA:
"Windows speed chart - times are given in seconds"

Firefox 1.0 (Moox):
20.33,2.78,3.18,1.57,26,2.84,41
Firefox 1.0:
11.54,2.52,1.81,1.48,23,2.05,41

Can anybody explain to me? The "unoptimized version" performs better than the optimized one?

O_o

You're right, obviously something's wrong here. Somebody please give the guy the REAL optimized version.

Re:One advantage to Firefox...

Probably because the MOOX author's benchmarks for evaluating performance look at the software as a whole rather than particular uses that can be isolated and improved. Also, some of the benchmarks seem a bit fuzzy ("dragging it into the browser window and measuring its load speed"). Especially when considering a performance difference of less than 5 percent. Why not disclose what the actual numbers were too? It would certainly help us evaluate how much human error is involved in the testing process!

The other half of it is that the builds essentially just set a few compiler options to use opcodes that may not be used (SSE2?) for webbrowsing. Additionally, its possible that some of the optimizations are hurting the cache with bloated low level code. It would be interesting to see if the Intel compiler provided any stronger oomph, at a pure compile configuration level. But we don't have any Intel CPUs in the house.

RTA

From the article:

The Moox Firefox install is actually slower than the standard Firefox versions distributed from Mozilla.org, even though it is supposedly optimised for my particular processor.

lynx

And lynx wins on the speed comparisons, hands down.

lynx...is there anything it can't do?

Re:lynx

lynx...is there anything it can't do?

Satisfy your porn addiction.

Re:lynx

lynx...is there anything it can't do?

Satisfy your porn addiction.


I beg to differ...
http://www.asciipr0n.com/ [asciipr0n.com]

Re:lynx

lynx...is there anything it can't do?

Render the tables in TFA correctly, re-sort the tables, etc.

Aside from its incredible speed, though, the best reason to use lynx is that you can keep it open in a little window on your desktop with nothing but text showing. Their motto should be "Lynx: It Looks Like You're Working!"

Re:lynx

lynx...is there anything it can't do?

You misspelled "can".

Re:I'll take

Perhaps you should take the opportunity to help address the problem.

I'm trying to get into the habit of sending letters to sites complaining that they don't work with Firefox or Mozilla. I figure if more people take their business away because they have useless IE-only pages, they'll be forced to revisit these ill-considered decisions.

The same thing is even more true for those sites with popup messages: "This site only works with Internet Explorer." Fix your damn site -- don't blame me for your stupid decision to hire VB programmers.

Hell, even microsoft.com is perfectly usable with Mozilla and Firefox. I certainly haven't noticed a "lack of richness in my browsing experience" there.

Re:Safari

You mean, way to go KDE, thanks Apple for contributing. Konqueror is not half bad, even if it's scripting speed is poor, as confirmed by the article.

Re:Closed Source Wins Again

Could the point be that Opera is closed source?

Question...

Is "not faster" a euphemism for slower?

To say that my camry is not faster than a porche 929 is a true statement when interpreted one way, but untrue when interpreted another. The use of amphiboly to lead someone to an erroneous conclusion is only different from an outright lie its craftiness.

Lee

Re:Question...

No offense, but I think you're using a lot of fancy words to tapdance around the (commonly accepted) fact that Opera is the fastest browser, followed by IE due to its native ties with the system, followed by Firefox because it reimplements all its own widgets in XUL, etc.

I really don't think there's much more to it. I use Opera on Windows specifically because it is faster and uses half the memory footprint Firefox does.

Re:Question...

I'm not really sure where people are pulling these stats. Probably from their asses, but when I load firefox, ie and opera this is the score:

P3 1GHz - 128MB Ram - Win2k Pro

Firefox loads the fastest
Opera loads almost as fast
IE... wtf is taking it so long if it's "integrated" as they say? It not only takes so much friggin time to load, but chews up the hard drive like they're going out of style!

Sorry. I'll believe my own results on the machines I use here.

450mhz 192Mb ram
500mhz 128Mb ram
1000mhz 128mb ram
2.8ghz 512mb ram
3.2ghz 2gb ram

all run win2k for games (sorry, new xp interface just doesn't cut it for me to mean a "new os") and linux for the main systems.

FF beats them hands down. I'm not a fan boy or anything, but it would be trivial to become one. I just use what's "WORKS" and works the fastest without pop-ups/problems/whatever.

-zo

Obligatory bash quote:

<kritical> matts: bikes go faster than cars...a bike at 60 mph is a lot faster than a car at 60 mph
<matts> kritical: um no...
<kritical> matts: um yes
<kritical> my sisters sport car at 60 mph goes faster than my dads explorer at 60 mph
<kritical> a bike at 60 mph will blow by a car at 60 mph

http://bash.org/?1988 [bash.org]

extensions

Interesting results. Firefox may be slower at rendering than IE and Opera, but I love the Firefox extension that disables auto-running flash elements in a page. For whatever reason, my work computer locks up on certain flash pages and this was a huge help.

This is really interesting.

I suppose the fact that IE has all sorts of nice direct access to the Windows code with god-knows-what tricks embedded to speed it up helps. Firefox is bound by what any non-MS program can do with the API.

That is not to say that I find Firefox slower - but thinking about it, I believe the Firefox interface (especially tabs and yes I know it was Opera first(?)) speeds _me_ up. So my perception is that using Firefox is generally faster than using Internet Explorer, even though it may be in actuality slower.

Really impressive work by that tester tho.:-)

Re:This is really interesting.

I suppose the fact that IE has all sorts of nice direct access to the Windows code with god-knows-what tricks embedded to speed it up helps. Firefox is bound by what any non-MS program can do with the API.

As far as "cold starts," keep in mind that 90% of IE loads into memory when Windows boots up, whereas very little of (e.g.) Firefox is loaded into memory. Really just the Windows libraries that it uses are loaded; all its own stuff has to load on the spot, but IE's rendering engine and various other libraries are all automatically loaded when the OS starts. That gives IE a huge apparent speed boost as far as starting it up for the first time after you boot the computer.

Re:This is really interesting.

"I suppose the fact that IE has all sorts of nice direct access to the Windows code with god-knows-what tricks embedded to speed it up helps. Firefox is bound by what any non-MS program can do with the API."

Nice try, but how does that explain IE being faster than FireFox under MacOS X as well in some areas?

Of course, Safari kicks them both :-)

Re:This is really interesting.

Nice try, but how does that explain IE being faster than FireFox under MacOS X as well in some areas?

Well, when you don't support entire chunks of the language you can be faster.

Speed tests mean nothing if the browsers don't render the results properly.

faster = better?

Just look at the Opera results for a moment. Notice how the later versions are actually slower.

But aren't later versions better, more capable, more adverse-effects resistant?

Also, a browser can render much more quickly if it doesn't care how badly it renders what you see. How does this balance with the loading times in the article?

ahem..

Can we get a realistic test? Lets see how quick IE is after a couple of days browsing some of the.... less family friendly websites. Firefox would rape it hands down.

Firefox patches

I recently switched to Firefox and on NTBugTraq last week, 3 exploits were announced with status of patched. I ran check for updates on firefox and reported nothing. I check A noticed [secunia.com] a bunch of other vunerabilities that say patched yet firefox.exe says there's no updates. I went to mozilla.org and even the default download is to the original 1.0 build. What gives? I'd expect update to actually work, there's no way i can install firefox on my parents machines because the only way they actually apply patches is when windows update actually downloads and prompts them. I can tell my parents to find the buried update feature and run it everyday, and that doesn't even seem to work.

I have to say...

Having deployed Firefox in a large installation, I noticed a great deal of complaints. While it seemed somewhat snappier, albeit slower to load, than it's IE counterpart, it was incapable of properly processing the internal helpdesk software that was designed with FrontPage to the latest standards.

Unfortunately, this meant rolling back to Internet Explorer. While I personally prefer Opera, most of the users agreed that Internet Explorer did the best at talking with the internet after this experiment.

Re:I have to say...

incapable of properly processing the internal helpdesk software that was designed with FrontPage to the latest standards

Excuse me, If I was Dogbert, my tail would be wagging right now.

You're designing your software with Frontpage?

Wow... that's great... There's your first problem.

Frontpage? Standards? What ones are those?

But which browser is faster...

...on returning the error message when the server is being pummeled by Slashdot readers?

Re:nice wording

Or one could say "Firefox beats Internet Explorer in the slowness category". That should pass the Slashdot test.

Quality

There doesn't seem to be an allowance for correctness of rendering and conformity of the javascript implementation. If you discard all requirements for rendering and outcome of the script, cat(1) is the fastest browser hands down. Which explains Opera's performance; Opera's rendering and scripting off by just the tiniest bit in every conceivable feature. There's a definite speed/correctness tradeoff and Mozilla has always opted for correctness when practical.

Mozilla faster than Firefox

From the article:

Surprisingly, Mozilla is now faster at most tasks than Firefox.

Again, I ask--what exactly is the point of Firefox these days? When it was being billed as the replacement for Mozilla's browser, it made more sense. But Firefox is neither faster or slimmer than the official Mozilla browser, and now it seems it's actually slower too!

I'm just curious what the incentive is supposed to be to use it over Mozilla.

Re:Mozilla faster than Firefox

I noticed that too and started to wonder. I can understand Mozilla being a bit faster than Firefox since they used the latest 1.8 builds, whereas Firefox branched off the trunk last summer. What really surprised me was that Mozilla beat Firefox in startup time significantly. That means either the Mozilla.org people did a hell of a job optimizing the startup time of the suite, or the extra complexity of the suite doesn't drag it down nearly as much as Firefox fans want to believe. I'm leaning towards the latter being the more significant factor.

Speed Not My Priority

For my browser choice, a few fractions of a second rendering doesn't make me feel warm and fuzzy. I get my cyber jollies from using a browser that has the least number of vulnerabilities [securityfocus.org]. Afterall, those few milliseconds don't add up to the all the down time you might otherwise be stuck with.

I'm a little dubious.

I tested browsers myself a while back with Stopwatch [numion.com], and I found Firefox to render consistently faster than IE6. I collaborated with others on the test, and we found that overall, Firefox was about 25% faster. There were some exceptions to the rule though... (most notably, mozilla.org rendered faster in IE. But Microsoft.com rendered faster on Firefox).

I honestly don't know what this guy did differently to achieve opposite results.

Speeding up Firefox the right way

This has been popping up on del.icio.us/popular for a while now:

Speeding up Firefox the right way [codebetter.com].

This page contains detailed tips about getting the fastest firefox experience, customized to different speed computers and network connections.

Whats the Point?

What is the point of this? I thought browser speed just didnt matter anymore, at least it doesnt to me. Does anyone even notice rendering anymore? I dont use a computer slow enough, nor have internet fast enough (only a T1) to notice any damn difference. This might have been interesting in the ancient slow days but anymore? come on?

And just how do you test a cold boot of IE? reboot the computer? And if your not using windows why would you ever shut off your browser?

The test is flawed. What are we measuring here?

Each intermediate page must be allowed to load completely ... This means that any indicators that the browser provides to show that the page is loading must show the page as loaded before navigating to the next page.

If you read this, you'll know that these benchmarks are mostly useless. How many people wait until a page is completely finished loading before looking at it or clicking links?

Users will tell you that Browser A "feels" faster than Browser B. This doesn't mean that A downloads and renders the entire page faster than B. It means that A displays the necessary content faster than B.

I don't care if it takes 2.5 seconds to load a page if I can see 75% of the content after 0.6 seconds.

Who cares when the progress bar disappears?

A few thoughts

There are a bunch of things I'd have done differently when doing a report like this.

The most important one is trying to measure something as close as possible to the Web browsing experience. That means loading pages over a network (at 56K, DSL, Cable, and/or T1 speeds, with some latency) rather than from local files, and loading pages that look more like a random sampling of Web pages rather than constructed examples (e.g., a page with tons of absolutely positioned elements). When the author of the test constructs examples like those used here for the "Rendering CSS", "Rendering Table", "Script speed", and "Multiple Images" benchmarks, the results will have a bias (relative to average performance browsing the Web) towards one browser or another. I'm not saying the author of the tests chose to bias it in a certain direction; merely that constructed tests like this will always have some bias. When such tests become widely used by the press (as iBench has), it even leads browser makers to optimize for the tests rather than for what matters for users.

Also, when testing startup times on Linux (especially cold startup), it makes a huge difference whether starting in a KDE (QT-based environment), GNOME (GTK+-based environment), or other environment, since it affects which shared libraries are already in memory. Testing Mozilla's startup times under GNOME (especially if using a GTK2 version of Mozilla under GNOME 2, or a GTK1 version of Mozilla under GNOME 1) would have improved its performance significantly.

Finally, Mozilla 1.8 hasn't been released yet, so I'm a little puzzled how it was tested. The released version will have changes from the current development version, so it will perform differently. It may be a slight difference, but the report should really say exactly what is tested.

Gotta love Opera

What I love about Opera is that it is fast with all of its features turned on. To have similar functionality on Firefox I need a dozen plugins that are not seamlesly integrated and that weight the browser down.

Still, for most people I recomend Firefox. Its lack of ads and free price cannot be beaten and its default feature set don't confuse people who switch fron IE.

Either way you can't loose. Its the only way to live malware free.

Cheers,
Adolfo

Speed after a few weeks use

Few people (mainly those in libraries/'net cafes, and privacy nuts) use a "clean" browser. Most people will have hundreds, often thousands, of links in their browser history, tens of megabytes in the cache, a big collection of bookmarks, and plugins like Flash and toolbars. In my experience, a browser will be nice and snappy fresh out of the box, but after a few weeks of piling these things on, it may slow significantly, either in its startup time or while browsing. Some browsers may be worse than others in this regard. The author of the linked article has done an outstanding job, but since it appears most of the tests were performed on freshly-installed, "clean" browsers, the results should be considered with caution.

reasons to use Firefox

it's free, and it's Free. Who cares if one browser is milliseconds faster or slower?

Safari and other Mac browsers

I'm actually a bit impressed with how well the 400MHz Mac numbers came out compared to the 800MHz PC numbers, that is Linux and Windows. Especially since they all had 256MB of RAM which everyone seems to say is not enough RAM for running OS X acceptably. The script speed seems to be the only dog for Safari. Perhaps this is something I should be mentioning to potential switchers.

Grain o' Salt

While these tests are nice for having empirical data, it's also important to not focus too much on this data. In many cases, the differences in results was not much more than a second. IE sucks for many reasons that are not its speed. Firefox and Opera have far more benefits other than what speed it displays pages at.

The point is, for most of these browsers, they all run 'fast enough.' A second or two here and there isn't going to significantly impact your browsing experience. Tabs, intelligent UI design, intelligent security decisions, and perhaps themes/extensions will add up to the overall experience.

Not really surprising to me...

IE is a slimmed down browser where I can imagine its rendering engine simplicity combined with Microsoft's unique experience with the Windows kernel and the integration makes for a fast browser.

Opera seems to be a minor miracle in terms of code optimizations, at least on the Windows platform, since it's not OS integrated or cheats with pre-loadings, and the Opera team lacks Microsoft developers with knowledge about undocumented API calls, etc. Still it usually beats IE hands down with a vastly superior rendering engine, on par with Gecko. It's only unfortunate it's ad supported and closed source.

Finally, Firefox/Gecko is a very nice open source browser with nice extension support, but building on the cross-platform UI toolkit XUL instead of using native widgets, along with being built for platform independence instead of being heavily optimized for various platforms (I imagine the Opera team has to do more work for their browser to work on other platforms). I think some of these things play a role in some of Firefox's speed issues. There's no problem with the code I think, just a side effect from what Mozilla is trying to accomplish with the code.

It would've been interesting to have him compare to K-Meleon or Galeon as well, since it's slimmed down to the bare bones Gecko layout engine with just minor stuff in addition, and that stuff is also using native widgets AFAIK. Might have a positive effect on the loading times at least.

Opera vs. Firefox

I would actually like to see a comparison of a P3 optimized Firefox (Moox) against Opera. My guess is that Opera uses speed optimizations for higher end processors that would not be available in the vanilla distro of Firefox.

Another question is, did they test the free "Adware" version of Opera or did they use the $40 "Commercial" version (I know Opera 8 was the Beta, so that one is obvious)?

I would personally like to see if Firefox could beat Opera with processor specific speed optimizations and some fairly standard performance tweaks to the about:config...remember, these optimizations would not be available on Opera...

I would also like to see how the much used Adblock extension slows down or speeds up Firefox in rendering some basic pages.

I wonder if they had pipelineing enabled in FF

I doubt they had pipelineing enabled on FireFox.

These days, I take tests like these with a grain of salt. Particularly after the Gartner groups speed tests of Windows vs. Linux. They tweaked the hell out of a Windows machine, and used a stock Linux install and claimed Windows was faster.

When called on it, they conceded.

I have a feeling something similar is happening here.

compare what's comparable.

Ok so I haven't read the article, but from the disclaimer, it doesn't sound like it's possible to make a fair test. What I mean, is that IE is EMBEDDED in windows. IE loads when you open windows explorer, or "My computer" or whatever else file-viewing window it's IE behind it. So there are no real "cold starts" for IE. So that's my first comment on comparing "cold starts" and "hot starts". Second, Firefox shows much more speed on a linux platform. I don't know if that's because I'm running gentoo with a bunch of USE flags to speed up and prelink on top of that, or if it's just because it's linux. Now on the other hand, there's no IE for linux (thankfully!!!). Besides, most users are concerned not about rendering pages but about connection speed and features of their browsers. Not the speed on the machine. Only at work or in a college dorm will you have a connection that could make those speeds perceptible to the user. So, next, comparing Opera to Firefox. Great. Whatever happened to the saying "don't look at gift horse in the mouth?". Opera is not free. Firefox is. Why would you compare something free with something you want a better quality from? It's fine if you want to determine whether it's worth spending the money on another browser, but then you're looking at features, not at speeds. After all, if the whole of the industry wanted lots of speed from their systems, they'd all have dual processor machines running a linux-smp enabled kernel, with blackbox only, right? So, while it may be interesting to compare the ALGORITHMS behind it all, it's not that interesting to me to compare actual speeds, because they're going to vary by environment, machine and user. Someone who has several apps open in the background will notice everything slow down a bit, when someone who only browses without popups will find it more responsive, at least for local operations. Just my $.02 worth.

Konqueror Conquers!

What's the fastest Free Software browser for Free Software operating systems? Konqueror! I can't believe this is being ignored in the summary. I can't believe this is being ignored by the posters. Except for script speed, Konqueror is faster than all other Free browsers on KDE. It's faster than every other desktop's native browser!

KDE needs to trumpet this one loudly. I think that stupid suggestion to replace KHTML with Gecko just died a quick and deserving death.