Firefox Nightly Beats Chrome in Speed And Webkit Nightly Eats Them For Breakfast.
We already knew that Firefox nightly beats Chrome in speed, the gap is getting wider with the latest Firefox builds (3.2a1pre). While Google Chrome uses webkit as a layout engine it uses it’s own javascript engine called V8. On the other hand webkit developers are quietly tweaking away its SquirrelFish engine for javascript speed increase.
Contestants:
- Google Chrome 0.4.154.29 running native on Windows Vista. Finished SunSpider JavaScript Benchmark in 1554.2 ms.
- Firefox Minefield 3.2a1pre running native on Ubuntu. Finished SunSpider in 1299.2 ms.
- Webkit nightly r38860, the latest build is only available on Mac OS X. Finished SunSpider in 1048.8ms.
While there will always be fanboys who prefer one browser over another (where the hell is my Opera on webkit?) - the one thing that we can agree on is that its a winning situation for everyone as all these rendering engines, javascript engines and, in the case of Chrome and Firefox, browsers are open source. The one with the largest browser market share is not anywhere close in terms of speed or standard compliance. Their dominance in market share will too not last for long.
Update: Should have mentioned before. All used the same hardware, a macbook core 2 duo with separate installs of vista and Ubuntu with bootcamp along with OS X of course.
Loading ...
Webkit is open source as well (LGPLv2/BSD-style) as it was originally based on KDE’s KHTML. Have a look at http://webkit.org/ . Though Safari is Apple’s closed source webkit interface, there’s a good bunch of open source webkit interfaces now (arora, epiphany, konqueror, shiira and some others)
Miguel, no where in the post does it say that webkit is not open source.
“all these rendering engines, javascript engines and, in the case of Chrome and Firefox, browsers are open source”
webkit = Rendering engine
yes, true enough. I hadn’t understand you correctly.
Why did you miss out Opera? I’d love to see how it compares to the others. It always -feels- faster to me.
Scott opera uses webkit engine. No point of testing it separately.
Opera doesn’t use the WebKit engine. Opera uses Presto.
Still doesn’t make sense.
Why do you single out Chrome and Firefox as open source, apparently making a dig at Safari, yet safari is never mentioned. Webkit is also open source, so all three engines that you’re comparing are all open source.
I just ran this test on and got 848ms.
I’ve never ran it before, reasons for disparity?
Note this was for the lastest Webkit Build on Leopard
Why would you make an attempt to compare three web browsers in three different environments?
Glad you wasted your time writing up something so useless..
Jason, glad you wasted your time being a useless hater.
This test seems deeply flawed. I would have expected to see the competitors compared running on top of the same OS on the same machine with the same available RAM, etc.
Unfortunately, these results really must be taken with a grain of salt. If only we could have these browsers on a single platform to run the sunspider tests. Windows should’ve been the platform of choice for the Webkit, Firefox, Chrome test. You probably could’ve squeezed in Presto as well, just as another reference. At least for me, usability trumps open-sourceness any day.
You ran your tests from 3 different environments… I dont trust your results, for example, linux is (presumably) more efficient at dealing with threads and performance that the other two oses you used. So, of course Firefox would come out on top. At the very least, all three environments deal with memory management and process management differently, so you should’ve tested them all in a windows environment (since thats the only OS that offers all three browsers). Until you get some better and more conclusive data, dont make such sweeping claims.
For everyone talking about the results and how they were achieved, note that at least the latest webkit stuff (as mentioned in the article) is only available for mac at this time. I do agree that a different platform for each browser is not the best way to run a test like this.
I’m just guessing you didn’t want Opera’s Presto to steal the show
Just as mentioned by several commenter, running the tests on different platforms and comparing them makes no sense at all.
For those wondering how opera stands against chrome, I’ve tested opera 9.62, opera 10.00 alpha 1 and chrome 0.4.154.29 on my laptop with vista x64, an amd turion ultra zm-80 processor and 4 GB of ram. Opera 9.62 finished the test on 6932.6 ms, opera 10 alpha 1 did it on 6335.0 and chrome on 1760.6 ms. Assuming firefox is better than chrome and safari better than firefox (performance-wise) then safari is indeed the best.
NOTE: I’m an opera user
Mandriva has a webkit package based on the nightly tarballs that I update every couple of weeks, and a Midori package built against it. You could use that to test a recent webkit on Linux.
It’s true as most people point out that to be comparative, the browsers *should* have been run on the same OS. To solve *some* of the problem with Safari, test the safari-version that *exists* on win, and then test that version against the latest beta on MacOSX, to get an impression. However, since all benchmarked in sunspider is the Javascript-engine, it should be allmost entirely CPU-bound, and OS shouldn’t matter THAT much.
I actually think it’s worse that the author thinks that javascript execution-speed is the same thing as browser-speed. No look at all into how pages are loaded, and how quick graphics really is. For instance, SVG we’ll hopefully make it’s inroads sometimes this decade, and it would be interesting to compare how they stack up in these regards.
In a very dirty test on my system (which is an up to date, i586 Mandriva Cooker installation) - I just ran each browser right now and ran the test, I have a ton of crap running in the background - I get the following numbers:
Firefox 3.0.4 (current Mandriva Cooker package) - 5474.4ms +/- 1.3%
Firefox nightly (http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-trunk/firefox-3.2a1pre.en-US.linux-i686.tar.bz2) - 2489.4ms +/- 2.8%
Midori 0.1.1 / WebKit 38707 (current Mandriva Cooker packages) - 1806.0ms +/- 2.0%
Opera 9.62 (http://www.microrpm.ca/pub/opera/linux/962/final/en/i386/shared/opera-9.62.gcc4-shared-qt3.i386.rpm) - 9733.4ms +/- 0.5%
I don’t know why Opera is so slow. There isn’t a single test that’s obviously wrong somehow, it’s just really slow at each test. Could be something to do with it being a Qt app and I’m running in GNOME, but I dunno.
To everyone complaining that Opera wasn’t included because it would have won:
Opera is going to get it’s butt kicked all over the place on this test until they come up with a way to run javascript as native code like all of the beta browsers above do. The engines above have all added that feature and it sped up their scores by an order of magnitude.
Opera 9.62: 5763.4ms
Opera 10 alpha: 4903.0ms
They also sped up their javascript engine.
Firefox 3.0.4 vanilla got ~13000ms…
Opera 9.62: 5763.4ms
Opera 10 alpha: 4903.0ms
They also sped up their javascript engine.
Firefox 3.0.4 vanilla got 12011.2ms…
I don’t know about all the “hi-tech” testing…but I do know that Chrome loads pages and runs quicker than Firefox or Safari on all three of my computers that are running XP. I use all three browsers at different times depending on my mood. My 13 year old daughter uses Chrome because like she said…”it’s just faster”.
New Opera 10 alpha rocks.
… umm yeah that is the universes worst benchmark ever… you are technically compareing apples to oranges here. ANYTHING on vista is gonna be slower than anything on OSX or ubuntu… the only way to really make this claim is if the test was done on identical systems with the same OS (or the same system with all of them installed)… thats like saying this porche is faster on the race track than this lambrogini on mountian passes…
Ah Ah ! your tests were made with a different OS for each test.
I got 3787.4ms on a 2.2ghdualcore firefox 3.0.4
…speed in numbers is always good but the real thing measures in user satisfaction…
i’ve been a long time firefox exclusive user..but then i tried chrome and it feel faster everytime i used it…i mean i still use firefox for many things but when it comes to check my daily websites i go to chrome every time
Webkit does do better in testing but 3.2a1pre minefield has a much snappier feel, and runs heavy sites much more reliably. And with a few about config tweaks, Firefox really blazes. Also webkit is missing a lot of features i have come to depend on in Firefox.
webkit is an engine, firefox is web browser.
Development of the layout engine now known as Gecko began at Netscape in 1997.
WebKit began in 2002.
WebKit began its development in 2002 as a derivative/fork of KHTML, which has been around since 1998.
Actually the first public release of khtml dates back to the end of 2000 and work really started a year before that. Before that it was khtmlw and not khtml.
But the fork part is right, apple chosed khtml as a strating for webkit over the other open source solution because the code is small, clean, standard compliant, easy to port and to support. (by comparison gecko’s code is large, heavy, dirty, so-so for standard, and terrible to support over time).
im testing out all of the nightly builds of each and i gotta say squirrelfish is amazing.
The Best Site to test Javascript is this one: http://www.andrewmin.com/webx/. chrome barely renders it. firefox 3.2prealpha gets it but with slow window dragging. webkit nightly r39370 is amazingly fast and currently the winner as of 12/22/08.