Linux Browser Shootout with Peacekeeper

Some time ago, I discovered a great tool from Futuremark called Peacekeeper. Peacekeeper is a browser benchmarking tool that gives you the ability to benchmark as many different browsers as you like, save the results and compare them to one another as well as some other common browsers.

Given some of the grumblings about Firefox performance on Linux, and the current debate about WebKit vs. Gecko, I thought it might be fun to take a handful of popular Linux browsers and see how they compare from a purely performance standpoint.

This is not a particularly complicated or scientific test… I just navigated to the Peacekeeper website and ran their benchmark in each of the browsers. The system on which I ran the benchmarks is a Core 2 Duo E7400 at 2.8GHz with 4GiB of RAM. It’s running an up-to-date Ubuntu 9.04 64-bit with the 2.6.30-9 kernel from Ubuntu Karmic.

Mozilla Firefox

Firefox performance on Linux is, as others have mentioned and proven many times over, horrible. My tests don’t show anything different. Here’s the performance breakdown for the Ubuntu packaged Firefox 3.0:


Firefox 3.5, which was just released today, provides an ever so slight improvement on the previous version:


If you take a gander at the Peacekeeper website, however, you can see that the Windows versions of Firefox are consistently faster than what I’ve been able to produce here. In case you were wondering, I did run both of these benchmarks in safe mode so no extensions would adversely affect Firefox’s performance. It doesn’t appear to have helped.


I’m not sure how popular Opera is, to be honest. I know of only a couple people who use it every day, though their dedication is borderline religious. Opera has never really been my cup of tea (if for no other reason than it’s built with Qt and I’m a Gnome user), but one of the arguments I always hear is about how fast Opera is. In my limited interaction with the browser during these tests, it felt very, very fast. Unfortunately, the numbers tell a different story.


Opera 9.64 posts numbers that are comparable with Firefox 3.0. The new Opera 10 Beta provides a marked improvement and actually manages to edge out Firefox 3.5 by 175 points:



Konqueror is probably the biggest surprise (disappointment?) in this whole round of testing. Konqueror uses the KHTML rendering engine, which is the foundation upon which WebKit is built. Given that, I was expecting Konqueror to open up a huge lead over Firefox in the performance department. To my surprise, Konqueror was only able to best Firefox 3.5 by about 100 points, not nearly the margin I was expecting.



Arora is a WebKit and Qt based browser that’s relatively new.


Looking at these numbers, I’m starting to wonder if maybe Konqueror and Opera’s less than stellar performance aren’t due to Qt. Taking into account the scores of some of the other WebKit browsers in this list, it seems like Arora should’ve posted much better numbers.

Google Chrome

Now that we’re moving into the WebKit based browsers, we can expect to see some more impressive numbers that should put into perspective how truly terrible Firefox performance is on Linux.


Being our first browser with WebKit underpinnings, Google Chrome does not disappoint. Chrome was able to more than double the performance numbers of Firefox 3.5, Konqueror and Opera 10 Beta (which is even more amazing when you consider those browsers are all running natively at 64-bit, where Chrome is 32-bit only at this point). Google’s browser made a huge splash on the Windows platform when it was released, and although the Linux port is still relatively immature, it is progressing quickly and is (at least in my opinion) suitable for everyday use.


I’m guessing most of you haven’t heard of Midori yet. Midori is a lightweight browser for Gnome built upon WebKit. Midori is insanely fast, just look at these numbers:


Midori scores almost 1000 points better than Google Chrome, a browser universally lauded for being blindingly quick. This browser, even in its current state, is a complete joy to use. I would highly recommend giving this one a shot.


Epiphany is somewhat of a testimonial of the performance of WebKit vs. Gecko. Some time ago, the Epiphany developers decided they would switch from using Mozilla’s Gecko rendering engine to WebKit. That switch has been a rocky process, but as the WebKit GTK port matures, so too does Epiphany (and Midori and Chrome for that matter). The performance implications of the switch to WebKit are nothing less than mind boggling.


In Gecko form, Epiphany comes in dead last. Its abysmal performance numbers land in the same ballpark with Firefox 3.0 (which makes sense, seeing how they use the same rendering engine), and that’s not a compliment. However, if you fire up the WebKit variant of Epiphany, something wonderful happens…


Epiphany-WebKit absolutely blitzes the Peacekeeper and annihilates all of its competition (even Safari 4 and Google Chrome on Windows), besting even Midori by over 200 points. Seriously, what an amazing achievement. I’ll definitely be paying close attention to this browser as it moves closer and closer to maturity.


Just to sort of wrap things up and drive home a few of the points we hit along the way, here’s a graph comparing all of the browsers mentioned today:


I think the moral of the story is pretty clear: Gecko cannot hold a candle to WebKit when it comes to performance. And mercifully, WebKit brings decent browser performance to those of us who have been suffering on Linux under the growing weight of Firefox. Though I love my Firefox extensions, I will happily bid them farewell in favor of the speed Midori and Epiphany are currently delivering.

9 Responses to “Linux Browser Shootout with Peacekeeper”

  1. FCom Says:

    Maybe the poor results of the qt browsers are because you are testing them in gnome. I wonder if they preform better in a kde environment, I would suggest testing them in Arch (kdemod) or mandriva (Kubuntu has awful performance compared to other kde distros)

  2. Dave Says:

    I need to clarify, and I’ll edit the post when I get a chance, but all the browsers were tested without anything else running, including a desktop environment. I did this by dropping to a console and doing ‘sudo /etc/init.d/gdm stop’ and then ‘sudo Xorg -ac & DISPLAY=:0 $browser’

  3. Lars Says:

    What make me wonder is, that if I Benchmark FF 3.51 under VirtualBox with XP it reaches 1.784 points, while the same FF Version tested under Linux reaches 1.669 points. The curious thing is, that the Linux-Version runs faster… why does Peackeeper come to a lower result?

  4. Dave Says:


    The only thing I can figure is that the “Complex Graphics” portion of the benchmark (which is the canvas tag) is what’s making the Linux native version finish before your virtualized one, but the Peacekeeper website says that particular benchmark is not included in the overall result.

  5. Paul Stenius Says:

    looks very interesting, i will be giving some of these a try. also i noted that google got the best score for “complex graphics” at 5845. im going to have to see what is actually considered complex graphics in the benchmark suite, but all the other metrics look good

  6. Midori is Rockin’ Fast | Qe2eqe Says:

    [...] [...]

  7. eukristian Says:

    I also heard about this test (actually this is how i stumbled upon your page) but I can’t finish the Epiphany test . It stops at Social networking , at filtering . I wonder what the problem could be . I doubt that the problem is with my system (Athlon X2 4600 + , 2 GB RAM , …) as I could finish the test with Firefox 3.5.6 (terrible 1343 points ) .

    But I guess your test gives me the conclusion (Thanks for that ) so I’ll keep my Epiphany browser .

  8. Gustavo Says:

    I have recent benchmarks for Firefox 3.6 Linux and Firefox 3.6 Windows both native and via Wine here, in case anyone is interested. The Windows system is a little undepowered in comparison and FF Linux still manages to come last. Here:

  9. the dsc Says:

    Weird. With me, epiphany, konqueror and midori were unable to finish the test. I didn’t try arora. Opera (now 10.60) had the best score, followed by chrome, and then firefox (iceweasel). Somewhat surprising, I’d think that chrome would lead the dispute by a reasonable amount. That’s because the “feel” of speed is opposite for the first two. Not to mention that opera seems to not be properly handled by the window manager or whatever. Sometimes I can’t focus again on some page item, do right click + mouse wheel to change the tabs, it simply does not respond until I select some window and then opera again. Flash on opera is also very problematic, I similarly can’t interact with it more than once. I seems that I have to right click in some html part every time I want to interact with flash, otherwise it won’t respond. That’s in debian, I must check how is the status on arch linux one of these days, perhaps it’s different.

Leave a Reply