After reading about the new JavaScript VM in Google Chrome (called V8) I was wondering how it would stack up against the new JS engines from Mozilla (TraceMonkey) and WebKit (SquirrelFish). I ran 3 tests across all the browsers, including IE8, and here are the results.
Setup:
- Chrome [Build 1583 - 0.2.149.27]
- FireFox 3.1 Nightly [1.9.1b1pre/200809020331]
- w/ javascript.options.jit.* set to true except in the SunSpider tests, where jit.content was set to false because it crashes. - WebKit Nightly [r36012]
- Internet Explorer 8 [IE8 v8.0.6001.18241]
- I doubt numbers are accurate, asked me once for Chrome tests if I wanted to stop the unresponsive script, about 50x for Dromaeo - Windows XP SP3, AMD Athlon 64 X2 3800+, 2GB RAM
- Restart each browser after every test
- SunSpider, V8 BenchMark, Dromaeo
Click continue for results and charts…
Bigger the Better
- Chrome - 1083
- WebKit - 191
- FireFox - 106
- IE8 - 42
Needless to say, Chrome’s V8 blows away all the current builds of the next-generation of JavaScript VMs. Just to be clear, WebKit and FireFox engines haven’t even hit beta, but it looks like the performance bar has just been set to an astronomical height by the V8 Team.
Update: Brendan has has posted new TraceMonkey favorable benchmarks with the latest TraceMonkey build vs V8. A little friendly competition is awesome but since all these projects are open source the community is the real winner here.
Also I downloaded the latest trunk build of FireFox [1.9.1b1pre/20080903034741] and re-ran the tests:
The latest build is definitely optimized for the SunSpider test, even beating V8. However, it still lags behind in Dromaeo and the Chrome Tests. (The V8 team optimized the engine specifically for the Chrome Test so it’s very biased.)
Update 2: Apparently Resig has been working on a v2 of Dromaeo that includes a lot more DOM manipulation tests.
- Chrome - 10800.8ms
- WebKit - 6456.6ms
- FireFox - 10119.8ms
- IE8 - DNF (Too many “slow script” interruptions)
Chrome and FireFox are evenly matched here. WebKit comes out on top…
Also, being the JavaScript ninja that he is, John Resig has a much better write up on the topic than I.
September 2nd, 2008 at 9:58 pm
[...] a cutting-edge, blindingly-fast open source JavaScript engine called [...]
September 2nd, 2008 at 10:41 pm
It certainly is an exciting time for front-end engineering
On a 2.6GHz MacBook Pro with 4Gb RAM, I get considerably better performance using Webkit r36013 (beware anecdotal evidence, heh). 1783.2ms for SunSpider[1], ~ 420-440 for the Chrome tests, and 883.60ms from Dromaeo[2]. /shrug
I wonder how much of that delta is attributable to platform/specs? The Windows branch of SquirrelFish has “room to grow”[3], and it is admittedly hard to test Chrome on a platform it hasn’t been officially released on…
Comparisons to IE are always good for a laugh, or at least a temporary masking of the abject hatred I have toward that eclectic epithet-eliciting, egregiously excremental excuse for a (browser) engine. /twitch
[1]: http://bit.ly/ssr36013
[2]: http://dromaeo.com/?id=26664
[3]: http://webkit.org/blog/189/announcing-squirrelfish/
(btw, copy and paste error in the Sunspider result links, FF can’t be _that_ bad…)
September 2nd, 2008 at 10:45 pm
So bit.ly just lost its place as my URL shortener of choice. Pity, cute name. The Sunspider results I attempted to link earlier are located here: http://is.gd/2ahh
September 2nd, 2008 at 10:46 pm
Hey Wayne, could you please try a TraceMonkey build from tonight, 8:30pm or so? We fixed the bug that crashed the date-format-tofte.js test in SunSpider, and other bugs. More in my blog soon on where TraceMonkey is at ~two months old, and where we are going.
/be
September 2nd, 2008 at 10:49 pm
@Daniel: thanks for the correction, the link is on a windows box at work so I’ll update tomorrow
@Brendan: will do first thing in the AM, I’ll need to run it on the exact same box as I ran the other tests.
September 3rd, 2008 at 2:24 am
I wish you could add Opera to the mix as well.
September 3rd, 2008 at 10:16 am
I ran a webkit.org - performance test myself yesterday at http://goit-postal.blogspot.com/2008/09/chrome-first-little-test-with-highly.html . It compares the overall time to previous tests with the same setup ( http://goit-postal.blogspot.com/2008/05/javascript-core-performance-in-actual.html ) so you can compare Google Chrome to FF2/FF3, IE7, Opera and Safari. Anyway, the results are pretty the same as yours ;- )
Just my 0,02$, Georgi
September 3rd, 2008 at 10:37 pm
[...] Chrome esce vincitore non solo nei confronti di Internet Explorer 7, Safari 3 e Firefox 3, ma anche nel testa a testa con la Beta 2 di IE8 e le versioni nightly di Firefox 3.1 e Safari 4. Va sottolineato come questi [...]
September 4th, 2008 at 12:48 am
I tested the latest revision of Webkit (build r36078), and got these results:
Dromeo V1: 808.40ms
Dromeo V2: 4647.00ms
SunSpider: 1539.8ms
V8 Benchmark: 523
From these results, if compared to the numbers on the graphs above, Webkit r36078 won on the SunSpider test, and Dromeo v2. It is almost half of Chrome on the V8 Benchmark, and 24.4% nearing to Chrome on the Dromeo v1 test.
September 5th, 2008 at 9:04 pm
[...] For more detailed explanation, please visit Wayne Pan’s blog. [...]
September 6th, 2008 at 5:18 am
[...] Wayne Pan e John Resig hanno pubblicato alcuni benchmark confrontando le prestazioni degli interpreti [...]
September 14th, 2008 at 11:48 pm
[...] interested in direct comparisons of Chrome, Opera and Webkit have a look at the work done by Wayne Pan and John Resig! Tags: browser wars, [...]
September 19th, 2008 at 2:52 pm
[...] Yesterday WebKit’s latest salvo is SquirrelFish Extreme. After chuckling a bit at the name, I decided to run it through the same benchmarks as I ran Chrome through. [...]
January 3rd, 2009 at 4:57 pm
[...] favorites” by not including Internet Explorer 7 in the list of alternatives, but JavaScript benchmarks clearly show that Firefox, Chrome, Safari, Opera (and many other browsers) are much faster than IE. [...]