In their Safari for Windows announcement, Steve Jobs and Apple marketing claim some serious performance advantages over IE and Firefox. My understanding is that they're using an outdated test suite that doesn't really measure real-world situations. These kinds of benchmarks are easy to cheat, too. A vendor could, for example, add code to a browser that reported to the test suite that the test was completed long before it actually was. A vendor could also implement a feature incompletely or fail to implement it at all, cutting way down on the time it takes a particular task or set of tasks to complete. One could certainly imagine a page rendering faster if images weren't loaded or divs weren't positioned correctly, or some other web feature were disabled, not implemented, or implemented incorrectly.
Now, I'm not saying that Apple has done any of this. What I'm saying is that these kinds of benchmarks can be misleading and can easily be manipulated.
So what's a user to do? My suggestion is try out the browser for a while and see how it fairs on the sites you regularly visit. If it's faster and functionally complete, great, but don't just take a vendor's word for it.
I use a lot of contemporary web applications from companies like Google, Yahoo, Zoho, and others and for me Firefox is considerably faster and more functional than Win Safari or IE 7. Michael Calore, at the Wired blog Compiler, seems to have compared the three browsers and his experience is also that Firefox is faster on Google services.
Have you done side by side comparisons? What's your experience with the mainstream browsers, Firefox, IE, and Safari?
Posted by: DH | June 12, 2007 9:05 PM
Not sure about load time, but I find Safari 3 notably more responsive than Firefox 2 when dealing with GMail and more particularly Google Reader.
Posted by: MK | June 12, 2007 9:59 PM
I always found Safari to feel faster. Before the unofficial G4 optimized builds of Firefox came out, I couldn't stand to use Firefox on my PowerBook G4.
That said, the few times I've had to use Windows in recent memory, Firefox has seemed very fast and responsive. So I've been under the impression that it's the Mac specific code slowing Firefox down.
Posted by: Eric | June 12, 2007 10:41 PM
I am not surprised that Firefox is faster when using Google services. After all, Google is a Firefox company. They use Firefox as their primary browser. They are obviously going to optimize their services for the browser they use in their daily work.
While Google services are indeed real-world, it's not really a fair measurement of the web in general, since the general web is certainly not optimized for Firefox to the same degree as Google services, for the most part.
So the interesting question here is which one is faster on "normal" sites where neither has an advantage because the site has been optimized specifically for that browser.
Posted by: Larim | June 12, 2007 11:46 PM
Thanks to all the security problems found in the first few days real world testing of safari is kind of dangerous. :-/ After all it possible starting an app klicking a link. And you can even pass parameters....
Posted by: Morty | June 13, 2007 12:35 AM
So far I only tested this small testcase
http://www.jeria.net/performance_test/
Safari gets the fastest test results on this one. Even faster than Opera (which usually has better JavaScript performance than Firefox).
Posted by: José Jeria | June 13, 2007 12:44 AM
In Firefox trunk, when I click "array and push method" in José's test, it reports that it took about 1.2 seconds when it clearly took several seconds for the table to appear. I think this is because Firefox waits until after the script finishes to do any layout. If you add "document.documentElement.offsetHeight" to the script before computing endTime, the results will be closer to reality. (You might still not be guaranteed a paint, though, just layout.)
Posted by: Jesse Ruderman | June 13, 2007 1:16 AM
FireFox on OSX is a joke ... it's way to sloooow (17" MBP).
Safari on osx is way faster then Firefox on OSX ... let's see how firefox will compare to a final safari 3 and/or other webkit/khtml/kjs browsers (gtk/webkit ... yay!) on win32!
Posted by: bustaa | June 13, 2007 2:20 AM
I tried out Safari on Windows a bit yesterday. I can't say I noticed any difference in load time or UI stuff, but browsing web pages did seem rather snappier than either Firefox or IE (as I also found with Opera last time I tried it out).
iGoogle and Google reader do seem to work rather better on Firefox, and so does the World of Warcraft armoury site. But then they are both designed with Firefox and/or IE in mind. In one way it's good that Firefox has enough of a usage share to get people to do that, but it still means sites are writing for IE and Firefox (as they used to write for (Netscape and IE), rather than writing for any browser.
Posted by: Michael Lefevre | June 13, 2007 2:55 AM
I browsed for a while with Firefox and Safari side to side and couldn't find any noticeable difference. I've tried this kind of benchmarking before with IE and Firefox, and as much scientific I try to get I always get to this:
- It depends what sites you visit
- The difference is hardly noticeable.
As much as I'd like to believe Calore's report (it would be great to have the fastest browser) it is obviously a handicapped "study": Google sites only, no Opera.
Web standards compliance, open development, a passionate community, extensibility and a better security record are solid statements that admit no rebuttal from anyone. And that's why I use Firefox and Mozilla products.
Posted by: Percy | June 13, 2007 5:59 AM
safari on windows is a JOKE, if apple want to commit suicide and let all the world see the real inability of their developer, I amn't gonna object it. :D
Posted by: cris | June 13, 2007 6:14 AM
safari on windows is a JOKE, if apple want to commit suicide and let all the world see the real inability of their developer, I amn't gonna object it. :D
Posted by: cris | June 13, 2007 6:15 AM
On my Mac (dual G5 powermac) Safari is noticeably faster after startup....have to bounce it once a day or it gets progressively slower as memory usage rises. This is Safari 2.0, not the beta.
On Windows it's tough to gauge. Rendering of many things seems lightning fast, and startup was fast, but it's subject to some hitches and pauses, which I could chaulk up to beta-ness. I'm guessing it will get really fast as it matures, but that's just a guess.
On the Mac, I use firefox for development (firebug!!!!) but Safari for day-to-day browsing.
Posted by: Rob Meyer | June 13, 2007 9:44 AM
Though Safari is a nice browser after all, I'm not convinced to use it on Windows, because it and its text-rendering just don't belong there. Mac-OS widgets on Windows suck, because the best skinning for almost any software is no skinning at all (with some exceptions in the media player field): software should respect the OS.
I don't think the speed of the browser itself shouldn't be overvalued, I didn't feel major differences between Firefox and Safari, and if, I think I would just put up with it, because browser speed mostly depends on download speed.
However, I'm a bit disappointed with how unreliable that so-called beta is ... it can't even bookmark a page without producing a crash and it doesn't render some fonts (so that you only see white space).
Posted by: Aaron Strontsman | June 13, 2007 12:38 PM
It runs so fast that it closes itself before I even think of clicking the X button! ;)
In all seriousness, these claims of "speed" remind me of ATi/nVidia designing drivers that look fast in benchmarks, but at the cost of the output being ugly (sort of like Safari/Opera's UI...)
Posted by: ant | June 13, 2007 1:32 PM
Here's a good look at how Safari handles certain kinds of speed tests.
http://www.howtocreate.co.uk/safaribenchmarks.html
via Opera Watch
Posted by: Tyrant | June 14, 2007 9:06 AM
Safari keeps crashing for me. Whenever I try to open sidebar, for example. I saw some other bugs, too. I think that it is too much for public beta.
Posted by: Ivan Ičin | June 14, 2007 1:10 PM
I did some real world tests of Safari vs Firefox based on the loading time for just visiting a range of "normal" pages (e.g. digg.com, google.co.uk etc - also included Safari and Firefox starting pages just for fun!)
My methodology was slightly different in that I used Wireshark to monitor the time taken between the first and last HTTP GET requests, but the results were largely the sameas other people's findings: Safari was on average about 20% slower than Firefox when loading pages in "real world" situations, although rather curiously it totally slaughtered Firefox when loading news.bbc.co.uk!
http://www.mdibb.co.uk/2007/06/14/safari-vs-firefox-benchmarking-real-world-performance/
Posted by: Matt | June 15, 2007 8:38 AM
I was about ready to post howtocreate's link, too. I like how in this entire article you managed not to say a word about Opera's speed (good thing you didn't because the only words you have for Opera are negative, even if sometimes Opera is faster). Nonetheless, I find Safari is a little faster than FireTurtle. Perhaps you could give us a couple positive words about this release, as it encourages people stuck on IE to become aware of alternative browsers.
Posted by: anonymous | June 16, 2007 4:57 PM
Ok, what about those web-services optimizing their code for certain browsers and making it clunky for other browsers? What's your opinion about that?