Subscribe to the feed.
Firefox 3.5 RC1 was released today. It is supposed to be a contender to Google Chrome which is probably the fastest browser currently.
Setup
Windows Vista SP1 Intel Core 2 Duo 1.80 ghz
Firefox 3.0.11
Chrome 3.0.187.1
Internet Explorer 8.0.6001.18702
Opera 10.00 Beta Build 1551
Safari 4.0 (530.17)
All extensions disabled. No 3rd party plugins installed besides Flash and Java.
SunSpider 0.9 Benchmark
http://www2.webkit.org/perf/sunspider-0.9/sunspider.html
Totals
Chrome 832.8ms +/- 1.7% Safari 833.4ms +/- 1.7% Firefox 3.5 RC1 1568.2ms +/- 0.6% Firefox 3.0.11 3945.0ms +/- 3.8% Opera 4048.0ms +/- 0.9% IE8 6671.8ms +/- 0.7%
Chrome vs Firefox 3.5 RC 1
TEST COMPARISON Chrome Firefox RC
=============================================================================
** TOTAL **: *1.88x as slow* 832.8ms +/- 1.7% 1568.2ms +/- 0.6%
=============================================================================
3d: *1.73x as slow* 136.0ms +/- 13.8% 235.4ms +/- 1.0%
cube: *1.70x as slow* 35.0ms +/- 4.4% 59.4ms +/- 1.1%
morph*: ?? 64.6ms +/- 29.5% 71.0ms +/- 3.0%
raytrace: *2.88x as slow* 36.4ms +/- 3.9% 105.0ms +/- 0.8%
access: *2.97x as slow* 65.8ms +/- 3.6% 195.2ms +/- 1.0%
binary-trees: *13.1x as slow* 4.2ms +/- 13.2% 55.0ms +/- 2.8%
fannkuch: *3.51x as slow* 24.8ms +/- 2.2% 87.0ms +/- 1.8%
nbody: *1.39x as slow* 26.2ms +/- 5.2% 36.4ms +/- 1.9%
nsieve: *1.58x as slow* 10.6ms +/- 10.5% 16.8ms +/- 3.3%
bitops: 1.06x as fast 55.0ms +/- 4.2% 52.0ms +/- 2.9%
3bit-bits-in-byte: 1.90x as fast 3.8ms +/- 14.6% 2.0ms +/- 0.0%
bits-in-byte**: ?? 11.2ms +/- 5.0% 11.4ms +/- 6.0%
bitwise-and: 4.71x as fast 16.0ms +/- 0.0% 3.4ms +/- 20.0%
nsieve-bits: *1.47x as slow* 24.0ms +/- 9.0% 35.2ms +/- 1.6%
controlflow: *13.2x as slow* 3.8ms +/- 27.4% 50.2ms +/- 2.1%
recursive: *13.2x as slow* 3.8ms +/- 27.4% 50.2ms +/- 2.1%
crypto: *1.80x as slow* 47.8ms +/- 8.9% 86.2ms +/- 3.3%
aes: *2.59x as slow* 18.8ms +/- 26.2% 48.6ms +/- 5.3%
md5: *1.60x as slow* 15.0ms +/- 5.9% 24.0ms +/- 0.0%
sha1: 1.03x as fast 14.0ms +/- 0.0% 13.6ms +/- 5.0%
date: *1.94x as slow* 122.8ms +/- 0.5% 237.8ms +/- 1.2%
format-tofte: *1.91x as slow* 62.2ms +/- 1.7% 118.8ms +/- 1.9%
format-xparb: *1.96x as slow* 60.6ms +/- 1.8% 119.0ms +/- 1.0%
math: *1.05x as slow* 79.4ms +/- 2.4% 83.2ms +/- 5.9%
cordic: *1.50x as slow* 27.0ms +/- 3.3% 40.6ms +/- 1.7%
partial-sums: 1.24x as fast 39.6ms +/- 2.8% 32.0ms +/- 13.2%
spectral-norm: 1.21x as fast 12.8ms +/- 4.3% 10.6ms +/- 6.4%
regexp: *4.87x as slow* 23.2ms +/- 17.5% 113.0ms +/- 2.3%
dna: *4.87x as slow* 23.2ms +/- 17.5% 113.0ms +/- 2.3%
string: *1.72x as slow* 299.0ms +/- 1.0% 515.2ms +/- 1.5%
base64: 1.74x as fast 40.6ms +/- 4.1% 23.4ms +/- 4.8%
fasta: *2.37x as slow* 46.0ms +/- 1.9% 108.8ms +/- 1.7%
tagcloud: *2.37x as slow* 54.2ms +/- 3.8% 128.2ms +/- 1.1%
unpack-code: *2.07x as slow* 95.0ms +/- 1.6% 197.0ms +/- 3.8%
validate-input: 1.09x as fast 63.2ms +/- 1.6% 57.8ms +/- 8.0%
* not conclusive: might be 1.10x as slow
** not conclusive: might be 1.02x as slow
V8 Javascript Benchmark
http://v8.googlecode.com/svn/data/benchmarks/v4/run.html
A benchmark to test Javascript performance. Widely believed to be heavily skewed in favor of Chrome. Larger scores are better.
Chrome 2377 Safari 1683 Firefox 3.5 RC1 279 Opera 182 Firefox 3.0.11 166 IE 71.3
Conclusion
Firefox 3.5 RC 1 feels much faster than the previous versions, and is catching up to Chrome. However, it is still about 2x slower than Chrome according to the SunSpider benchmarks.
Related posts:
I don’t understand why you didn’t include Firefox 3.1 in your benchmark. How can you claim it’s faster when you can’t even be bothered to include the old version in the benchmark for comparison?
I only said it felt faster which is obvious to anyone using it. However, I have put in the results.
Please also include the original webkit (without google’s v8 javascript engine).
Mozilla plan to trace recursion and DOM operations, that would make them much faster. We’ll see.
I included Safari. Is this acceptable?
Hasn’t tracemonkey been included in FF since 3.1? I thought it had, I may be mistaken. Hotpath tracing isnt always a performance win though.
Running a bunch of created tests by webkit and v8 (which are the underlying technologies of Chrome)… can we have some neutral tests please?
Can you suggest any?
This Mozilla blog uses the SunSpider Benchmark as well:
http://weblogs.mozillazine.org/roadmap/archives/2008/09/tracemonkey_update.html
I don’t think it’s reasonable to say that the Chrome tests are “heavily skewed in favor of Chrome”. They’re well-known and widely-respected old Lisp benchmarks. It’s more accurate to say that Chrome is heavily skewed in favor of those benchmarks.
JS benchmarks are interesting, but it would be great to see some DOM manipulation comparisons too. My observation has been that Webkit based browsers are significantly faster at rendering than Gecko.
Thanks for posting your comparison, I installed 3.5RC1 but didn’t notice a significant difference then immediately regretted not running a proper benchmark to compare the two. So basically 3.5 is around 2.5x faster (250%) than 3.0.x but that’s still half the speed of Safari 4 or Chrome so it’s still slow in comparison.
I was pretty disappointed because I didn’t notice the difference in normal web usage and I’m shocked to see a benchmark shows the JS speed to be that much faster and I can’t notice it in real world usage.
PS: thanks for posting this, it needed to be said because all I see from the other commenter is whining. If they want some niche benchmark comparison, then they need to run it them self, running 3.1 is just silly you’re comparison was extensive.
It would be awesome if we at one point started benchmarking how browsers run *applications* such as gmail, google docs etc… That would be meaningful, and much more in line with how we benchmark other technologies.
I dont believe this article, Opera works better for me, and got something called functionalty, FF needs extensions to do that.