Some conclusions:

• Firefox is over-represented: although Firefox has a good 20%+ or so world-wide market share, the 52%+ result here indicates that Firefox users are open to new browser technologies (no surprise there)
• Safari: ditto, although I suspect that some of the Safari visits are indeed Chrome visits before Google Analytics got smart enough to recognize it
• Chrome and Opera: Chrome got even with Opera in two and a half days. That is a statement, and might get Opera people a bit worried about market share.

Stop the presses!

A few hours (like 6) later we have a somewhat different picture!

Chrome gained 7%+ and it seems that my Safari-Analytics-Chrome theory has some substance as Safari is down markedly, while the other players are virtually unchanged.

Day 4: Chrome is second

So we are at day++, and the results are:

Chrome again gained 7%+ and is second only to Firefox, slightly edging out IE. Not bad for a four day older isn’t it?

I run three benchmarks (on the same machine, all browsers open, running the benchmark only in one at a time):

• the V8 Benchmark Suite
• a nice SVG & Canvas speed test
• Sebastien Gruhier’s PGF test page (used earlier in a previous test).

The first test is more JavaScript oriented, where the new V8 virtual machine can shine, the other two test graphical abilities.

And the results are…

The V8 Benchmark Suite results are:

• IE 7.0.5730.11: Score: 16 (Richards: 10; DeltaBlue: 8; Crypto: 23**; RayTrace: 20; EarleyBoyer: 31)
• Opera 9.52 (10108): Score: 102 (Richards: 57; DeltaBlue: 73; Crypto: 65; RayTrace: 142; EarleyBoyer: 290)
• Safari 3.1 (525.13): Score: 59 (Richards: 35; DeltaBlue: 45; Crypto: 50; RayTrace: 81; EarleyBoyer: 116)
• Firefox 3.0.1: Score: 77 (Richards: 78; DeltaBlue: 88; Crypto: 56; RayTrace: 65; EarleyBoyer: 111)
• Chrome 0.2.149.27: Score: 676 (Richards: 858; DeltaBlue: 777; Crypto: 650; RayTrace: 345; EarleyBoyer: 946)

There were significant (cca. 10%) fluctuations in each result, but it is obvious that in this (biased) benchmark Chrome is at least 6.7x faster than the next best, and 42x faster than IE.

The SVG & Canvas results:

• IE 7.0.5730.11: does not support SVG & Canvas, so did not complete
• Opera 9.52 (10108): SVG: 1.657s; Canvas: 1.75s (both beautiful)
• Safari 3.1 (525.13): SVG: 4.312s (edges clipped); Canvas: 1.5945s (beautiful)
• Firefox 3.0.1: SVG: 1.46s; Canvas: 1.095s (both beautiful)
• Chrome 0.2.149.27: SVG: 2.877s (line aliasing and clipping problems); Canvas: 0.993s (beautiful)

Again, there were some fluctuations, but we can conclude two things:

• Canvas is king in speed, and in Canvas Chrome seems to have a very slight edge over Firefox
• Canvas rendering quality is consistent and top notch among the browsers; SVG rendering quality is not consistent

The PGF test page results:

• E 7.0.5730.11: VML: 3390ms (both text and image are missing)
• Opera 9.52 (10108): SVG: 1282ms (image is missing on the right); Canvas: 485ms (both text and image are missing)
• Safari 3.1 (525.13): SVG: 625ms; Canvas: 328ms (text is missing)
• Firefox 3.0.1: SVG: 630ms; Canvas: 385ms (text is missing)
• Chrome 0.2.149.27: SVG: 593ms; Canvas: 281ms (text is missing)

Here Chrome seems to edge out the other browsers slightly, and IE is terribly slow (like it used to be). It is also plain that Canvas support is not complete yet in any of the browsers.

More later…

So the browsers are:

• Internet Explorer 6.0.2900.2180.xpsp_sp2_qfe.070227-2300 :-)
• Firefox 2.0.0.4
• Safari for Windows 3.0.2 (522.13.1)
• Opera 9.21 (Build 8776)

I run the page several times in SVG, VML, and Canvas mode on each browser (of course, when it was possible) with the following results:

• Internet Explorer: only VML is available
  • VML: originally around 4300ms, and with each re-rendering the time increasing by cca. 1800 ms (!) so it went like 4300, 6xxx, 8xxx, 9xxx, and finally jumped to 13109 ms
• Firefox:
  • SVG: generally around 1900 ms, sometimes dropping to 1500ish or peaking at 2300ish ms
  • Canvas: odd runs: ~1100 ms, even runs: ~750 m, alternating :-)
• Safari for Windows:
  • SVG: dead steady on 1070-1030 ms
  • Canvas: pretty steady around 600-468 ms, mostly in the 490-468 ms range
• Opera:
  • SVG: pretty stable at 2360-2078 ms
  • Canvas: pretty stable at 1030-860 ms

Again, the runs were performed on the same machine (2GHz P4, 2GB RAM, WinXP Pro, Commit charge ~1369MB), all browsers running at the same time, but the test performed sequentially.

My informal results:

• SVG: Safari is the king, at least 50% better, but typically 90-100% better performance than the competitors
• Canvas: again Safari is #1, at least 25% speed advantage, but as good as 100% better — with quite stable performance in both cases.
• VML: well, let’s read between the lines ;-) (please note: I did not run IE 7, but IE 6, for technical reasons) – if we consider its best result in VML and compare to Safari’s average SVG result (VML and SVG are closer relatives than VML and Canvas), then Safari is 400% faster.
• Canvas is the overall spead king, SVG is #2, and VML is distant #3.
• Safari is king in graphics — and this is still the first beta release from Apple…

BTW I got into all this because I’m looking into how to best implement some of my stuff on the iPhone with Morfik, and also for some graphical stuff I plan to do (charts, Gantt, diagrams, hint-hint:-).

Again, this is not science here: I only tested one rendering task, the environment was quite loaded, did not test IE 7 (although the rest of the browsers were their latest), did not check rendering quality in depth (things looked pretty OK in all cases), etc. Still I consider the results quite relevant — I’d like to see what other pepole can say about their experience…