Suspected PayPal Hackers Seek Deal To Stay Out Of Prison
2 Men Arrested After Teen Was Run Over For His iPad
4-way HTML5 speed test: Firefox 3.7 faster than Internet Explorer 9 (video)
As you will see from the video, Chrome (even with hardware acceleration turned on via command line switches) is in a distant last place, miles behind both Firefox and IE. Opera competes well up to a point, but eventually gives way to Firefox and IE with 1000 fishies swimming about.
I state in the video that Opera is hardware accelerated, but I could be wrong (we've confirmed that they have hardware acceleration devs on staff, but don't know what's under Opera's hood just yet). Chrome also performs very poorly, even though it (apparently) has hardware acceleration. Incidentally, if you want to turn on hardware acceleration in Firefox 3.7, follow this guide.
Both Firefox and IE9 use Direct2D to utilize the GPU's rendering power -- so it's probably no surprise that they both show very similar results. It's most apparent when comparing CPU use; IE9 and FF3.7 are both miles ahead in performance, but both show the lowest CPU utilization!
Still, I'm certain Chrome will feature full hardware acceleration soon enough. The thing I'm most interested in is whether we'll see cross-platform hardware acceleration. Direct2D doesn't exist on either Mac or Linux -- so we'll see how it pans out! Meanwhile, if anyone wants to pit Opera or Chrome against Firefox on Linux, I'd be very interested to see the results.
Comments
49
Subscribe to comments216Jun 24th 2010 10:20AM
"Chrome (even with hardware acceleration turned on via command line switches) is in a distant last place, miles behind both Firefox and IE. "
Its simply amazing how quickly things change in the browser wars
F-ZeroJun 24th 2010 10:24AM
i'm not an expert at this in any way, but shouldn't you be doing this on 4 separate identical computers? (so each programme can be allowed to maximise resource usage?)
SachmoJun 24th 2010 10:38AM
That would definitely make for a better experiment, but I think he was just showing the huge differences in these developer releases. Undoubtedly Chrome 6 performs worse on this test.
The Windows NT scheduler should do a good job of switching between the four, but I'm not sure how it behaves with Direct2D in the mix. I wouldn't be surprised if it gave priority to Direct2D applications.
Sebastian AnthonyJun 24th 2010 11:18AM
Ya, I did think about that -- but don't worry, even if you run Chrome on its own, it's still that slow.
If you look at the Task Manager half way through the video, you can see my CPU isn't above 40-50% or so (and there's plenty of free RAM). It's not hitting the hard disk at all.
I think Firefox saw a slight increase in speed (1-2 FPS) when I closed IE9, but Chrome was definitely just as slow on its own, or with the others open.
arashJun 24th 2010 6:28PM
yeah i tested one browser at a time on a t7300,8400gt,2gb system and ie9 was slighty faster that 3.7 webm with enabled h/a,for ie with 1000 fish it was 7fps and for ff3.7 1fps.i didn't see any major difference in other tests between ie9 and ff3.7 with h/a. ie8,chrome,opera,and firefox without h/a are dead officially
SevenmackJun 24th 2010 10:52AM
Well, maybe he should use four different computers. But in the real world, you may have IE and Firefox open on one computer. I have three browsers (including Opera) open on my main laptop right now. So the approach Sebastian has taken likely makes the best sense given what we actually do in the real world.
TheOneAndOnlyJHJun 24th 2010 11:52AM
By saying it "makes the best sense given what we actually do in the real world" you're referring to a small crowd of 'we'. Most people I know don't use three browsers at once, they just install the one they think is best at the time, and continue to use it solely from then on, regardless of the latest developments. Even if they install a second browser, they probably use it only rarely (such as on some interactive sites for banks or college courses that only run in 'supported browsers').
Personally I have four different browsers, yet I still use Seamonkey2 most. It may not be as fast as some others, but I use the integrated mail options, and I really like the advanced features and menu layout and ordering (even better than Firefox). I have used other browsers that really are faster, but not by enough to make me switch. I'm more concerned with the interface than those few hundredths of a second.
venom8599Jun 24th 2010 11:17AM
If you were still using the current Chrome 6 developer release (6.0.437.3) for this then you may have a small problem. If I remember correctly, the command line flags to enable the hardware acceleration are only supported in the last few nightly builds of Chromium; they haven't yet made their way into the Chrome Dev releases.
Sebastian AnthonyJun 24th 2010 11:18AM
Oooh, that's a good point.
I'll check -- and if I'm using the wrong version, I'll give it another go :)
Sebastian AnthonyJun 24th 2010 12:09PM
Tried with the latest build (50721) and it doesn't seem any faster. The meter also goes up and down, like in the video -- if that's indicative of anything :)
venom8599Jun 24th 2010 3:57PM
I performed a test of my own, a little earlier. Using the flags for me had no effect on the current Chrome 6 Dev build, but I was getting a much slower 3-4 fps average frame rate for 20 fish. Using Chromium build 50721, however, my results were identical to yours, though it appears I didn't need to use the command line flags to enable the acceleration, as I got the same result whether they were present or not. It was at least some hardware acceleration, but it looks like the code could use quite a bit more polishing.
Thanks for replying and looking into this, by the way. :-)
Sebastian AnthonyJun 24th 2010 4:03PM
No problem! I'm well aware that this fast-and-furious benchmark is far from accurate, but I'm more than willing to make sure it's as fair as possible :)
digitalseditionJun 24th 2010 11:26AM
OpenGL exists everywhere and for 2D hardware acceleration (especially for the bits they need) it will be very fast and consistent so it should be pretty straightforward to get cross platform hardware acceleration for everything. I'm actually surprised that people hadn't been getting a GLContext and drawing to that at this point in the game.
Sebastian AnthonyJun 24th 2010 12:08PM
Ya, good point. I can only presume they'll be using GL for acceleration on Mac/Linux.
I also don't know if GL has an equivalent to Direct2D -- but I'm no GL master :)
Bas SchoutenJun 24th 2010 1:24PM
... Except that you'd need to write something to tessellate the vector graphics, and write anti-aliasing algorithms (Direct2D does much better than the MSAA 8x on your GPU).. And then you'd have to write code to map images to complex paths to do things like bitmap brushes... No, OpenGL does not provide what Direct2D provides. It may provide what Direct3D provides(but even that not completely), but it is no tool yet for cross platform vector graphics rendering.
octoberasianJun 24th 2010 1:38PM
There isn't a Direct2D equivalent for OpenGL. From what I'm reading, it's more complicated to do so since you still have to work within a 3D environment by setting up the viewport to a 2D plane. You can think of it like those 2D-style games but with 3D graphics.
Other possibilities is to make a custom 2D engine within OpenGL or use third party APIs to get 2D through OpenGL. The one API that comes close at least is Java2D which has hardware acceleration through OpenGL. The other is cairo.
If I let my imagination run about, the next possibility is programming something akin to Windows Aero GUI that uses Direct3D for the windows on the desktop, yet still retain a 2D look. But, it still forces OpenGL onto a 2D plane regardless.
As much as I want to see hardware acceleration of 2D graphics for OSX/Linux equivalents of Firefox and Chrome, I just don't think it'll be available at launch when these new versions come out. It'd probably take the OpenGL developers, Khronos Group, to update OpenGL API to include native 2D hardware acceleration and rendering.
One can only hope, right?
Jason DashJun 24th 2010 11:59AM
Opera isn't hardware accelerated at the moment - it's software accelerated for now, although they designed their new Vega graphics engine with the thought of adding hardware acceleration in mind, so it shoul be hitting in Opera's next major version, probably V11.
(Check this post, under Vega http://labs.opera.com/news/2009/12/22/ )
Sebastian AnthonyJun 24th 2010 12:07PM
Ah! That makes sense -- and also interesting that Opera's software rendering is thus more powerful/optimized than Chrome's...!
SilverWaveJun 24th 2010 12:34PM
So jam tomorrow?
Just when will this be out of preview and in use?
2years 3years anyone?
SilverWaveJun 24th 2010 12:38PM
hmmm "The final build of Internet Explorer 9 is expected to be released in 2011. "
We will see :-)