I discovered the SunSpider benchmark through a Jeff Atwood’s post on the same topic where he has given the JavaScript performance comparison for different browsers. His list do not include the browser betas though. Since I’ve been trying out Firefox 3 beta, and it seems a lot snappier than version 2, I decided to give it a try. To make things interesting, I also tried out the benchmark on the other browsers I had – Opera 9.20 (not the latest), Flock 1.0.3 (which is based on Firefox 2) and IE7.
I ran the benchmark once each on the browsers. My system configuration is as follows:
- Pentium 4 2.8 GHz
- 2 GB RAM
- Windows XP Professional with SP2
Do note that I did not create any special setup for the browsers, i.e., they had their plugins and extensions enabled. This might have affected the performance, but then again it also simulates more of a real world scenario.
Results (links lead to results on SunSpider page – complete results in this Google spreadsheet)
| Browser | Time (ms) – lower is better | % better than FF2 |
|
37066.0 |
0.0 (baseline) | |
|
18250.4 |
50.8 |
|
|
16815.0 |
54.6 |
|
|
32145.8 |
13.3 |
|
|
88694.4 |
-139.3 |
Observations
The benchmark result shows that there have indeed been significant improvements in Firefox 3 over 2, with an overall gain of over 50%, and the performance is close to that of Opera. There are tests where Firefox trumps Opera, but overall Opera still has the lead. However, there seems to be some downsides to the Firefox 3 optimizations at least as of beta 2, as the rich interfaces of Yahoo Mail and Windows Live Mail do not work.
The other important observation is that Flock is significantly faster than Firefox despite being based on the Firefox 2 code. The Flock developers have definitely optimised the code, and this shows. (Update: not sure about this, based on Mike Shaver’s comment below, and my own observations on a different PC)
As for IE7, it actually performs reasonably well, and is on par with or better than Firefox for a majority of the tests (also shown by Jeff Atwood). However, the string tests turns out to be IE7’s Achilles heel with the time taken (72347.8 ms) being 10-15x that of the other browsers.
I haven’t tried out the benchmark on Opera 9.5 (still in beta) yet, and there may be further optimizations there too. Safari was also pretty fast as per Jeff Atwood’s post. I hope to test out these two browsers as well, and possibly on other systems to see how much the system configuration affects the results.
Check out the detailed results in this Google spreadsheet (Firefox 3 seems to have problems with Google Docs – it was only showing the all documents page).
Update: Based on Mike Shaver’s suggestion (comment no.1) I tried out the tests for Firefox 2 and Flock on a different PC (Athlon 64 3200+, 1 GB RAM), both in safe mode and with extensions enabled, and found the results to be similar (FF2 is slightly faster this time). I hope to do more tests, and in the mean time if you have any results or observations to share you are welcome to share them.
| Browser | Time (ms) – lower is better |
| Firefox 2.0.0.11 w/ extensions |
29638.0 |
| Firefox 2.0.0.11 safe mode |
28338.0 |
| Flock 1.0.3 w/ extensions |
30496.4 |
| Flock 1.0.3 safe mode |
29098.6 |
AB's Reflections 
The Flock guys will tell you, if you ask, that they haven’t done anything on the core (and that if they did it’d be upstreamed to Mozilla’s CVS, naturally). I had just this conversation with them quite recently!
I think it’s much more likely that differences in configuration — especially the presence of different extensions — account for that performance difference. I don’t think you’re really saying anything about “why Flock is faster than Firefox 2”, as your title claims, when you take the browsers with different extension configurations intentionally. Even just running them both in Safe Mode would give you a better picture of the performance difference, if any; I’d be quite interested to see such a followup!
I don’t think any of the JS engine improvements are related to the issues with Yahoo Mail and Windows Live Mail; there’s at least one fix on the Yahoo side that will be pushed out soon, AIUI, but in both cases I think the problems are related to a change we made for better DOM specification conformance. If there are JS-engine changes that broke any web page, we haven’t heard of it, so if you’re seeing otherwise please do report them through our bug system (or, if for some reason you can’t do that, just send me mail with the details).
Thanks!
Mike
Hi Mike,
Thanks for pointing out the Flock development part. It is quite possible that the differences in extensions is creating the problem (there are quite a few of them in common installed on my machine though). I have found Flock to be more responsive than Firefox 2 and I know of a few persons who have switched to Flock for this reason.
My results are not that objective as I have noted, and I did think of running the tests in safe mode. Let me see when I can do that.
Another thing of note is that the differences reported by Jeff are relatively smaller, which can be explained by his much more powerful setup. I was also wondering whether we could check out how much impact the CPU has on these tests. Maybe the guys at Anandtech or Techreport might be of help.
As for the Yahoo and Live mail problems, let’s see whether the next beta/RC fixes that. My bit on the tweaked JS engine resulting in the problems was merely speculation, in the absence of other details.
This post is very hot, it is high ranked at http://www.adminor.info (daily weblog, weblog post ranking site)
Flock is faster than Firefox2? thats surprising! 🙂
The other interesting thing is that Opera still beats Firefox 3 beta2..