Linked by Thom Holwerda on Tue 16th Mar 2010 16:54 UTC
Internet Explorer As predicted, more Microsoft news from MIX10, and this is some big stuff: Internet Explorer 9. As we all know, Microsoft really let Internet Explorer rot away, allowing competitors to make much better browsers with better standards compliance and performance. With IE9, Microsoft is aiming to not just close that gap - but to overtake the competition. Update: Ars has an in-depth look at the platform preview.
Thread beginning with comment 413879
To read all comments associated with this story, please click here.
SamAskani
Member since:
2006-01-03

Just a quick test

http://ie.microsoft.com/testdrive/Performance/01FlyingImages/Defaul...

Ubuntu 9.10 x64, Nvidia 8400 GS, Nvidia drivers 195.17 (those are the ones required for CUDA Beta 3.0), Xeon E5405, 8 GB

Under Google Chrome 5.0.342.3 dev it runs miserably (4 fps with the default image size )

but with Firefox 3.5.8 it runs quite smooth (55 fps with the zoom in activated, no matter how fast is the spinning)

The pic in the blog shows 60 fps and it is supposed to be GPU-accelerated. So 55 FPS under Linux looks quite good.

I just took a quick look of the code and it is just javascript scaling the pictures so this looks more a test of how fast the javascript machine and the render system can work together.

I'm wondering how Firefox achieved such good performance, my understanding is that 2D acceleration in Firefox is not yet implemented under Linux, in principle 2D acceleration is in the pipeline and this only for Windows.

My guess this is due only to javascript performance to render the whole scene before sending to the graphic subsystem.

Another interesting test, using Vmplayer to run Win XP 32 bits and Firefox 3.6 shows stills a good 40 fps. And for this one the rendering passes several layers between Windows, vmplayer and finally Xorg. IE8 shows only 3 fps and Chrome even less with 2 fps.

I wonder then what exactly are we looking on the 60 fps shown with IE9: purely GPU acceleration or just the improvement of their javascript subsystem?. Anyways, this is a very simplistic example but still interesting to see the degree of performance of the javascript engines to render an scene composed of few flying images.

Reply Score: 2

Thom_Holwerda Member since:
2005-06-29

Under Google Chrome 5.0.342.3 dev it runs miserably (4 fps with the default image size )


Same here (on Linux too). Remarkable how much Chrome sucks on that test.

Reply Parent Score: 1

Bill Shooter of Bul Member since:
2006-07-14

Here's the odd thing (under linux of course):

Chromium 5, 5 fps
Konqueror 4.41, 40 fps


That's just messed up.
For those interested:
Opera 10.10 6 fps
firefox 3.58 49 fps

Reply Parent Score: 2

fithisux Member since:
2006-01-22

A naive question is How SVG rendering is GPU accelerated? I think only the canvas operations are accelerated. I also believe that 2D acceleration is necessary anyway to have a good experience n the desktop (I hate my VESA supported V3 XGI, have to test xf86-xgi). So 2D acceleration of canvas != GPU acceleration. Am i right?

Fianlly I hope to see a DirectX/OpenCL/OpenGL whatever API at the hardware level. If GPU rendering will become the norm OSes must have it anyway with specs and standards, without proprietary drivers, just like VESA.

Reply Parent Score: 2

Neolander Member since:
2010-03-08

"Fianlly I hope to see a DirectX/OpenCL/OpenGL whatever API at the hardware level. If GPU rendering will become the norm OSes must have it anyway with specs and standards, without proprietary drivers, just like VESA."

Not sure that I agree with that.

Norms like VESA appeared because at the time, it was thought that hardware should offer advanced capabilities to all devs, get compatible with software.

But times have changed. Now, for the sake of reduced hardware price, the trend is that a lot of hardware (including very widespread one, like wireless) would use proprietary technology instead of norms, and that the provider would provide a driver for Windows only.

3D acceleration is already everywhere. It is on more than 95% of the computers due to Windows Vista/7 specs. I don't think pushing it forward even further and making it mandatory for normal internet operation will make standardization less painful. Instead, it'll probably make thing worse for indie OS devs who only get beta-quality driver or no driver at all from the manufacturer.

Edited 2010-03-16 20:35 UTC

Reply Parent Score: 1

google_ninja Member since:
2006-02-05

hardware rendering is a completely implementation specific optimization, specs and standards for it would be silly.

Reply Parent Score: 2

umccullough Member since:
2006-01-26

I wonder then what exactly are we looking on the 60 fps shown with IE9: purely GPU acceleration or just the improvement of their javascript subsystem?. Anyways, this is a very simplistic example but still interesting to see the degree of performance of the javascript engines to render an scene composed of few flying images.


With intel graphics on winxp, the FPS varies quite a bit in FF 3.6 depending on where you have the mouse located in relation to the images, and how fast they're moving. I can get it to go as low as 20fps, and as fast as 66fps.

And yes, Chrome 4.x performance is horrible here as well.

Reply Parent Score: 3

dinamic Member since:
2010-03-16

Ubuntu 10.04a3 Amd Athlon 2500+ Nvida 7600GS Firefox 3.7 a4pre => 47 FPS

Reply Parent Score: 1

j-kidd Member since:
2005-07-06

With a crappy Intel C2D T5550 plus an even crappier Intel GM965, I can get 60 FPS from Opera 10.51, Firefox 3.6, and Konqueror 4.4.1. Only Firefox has noticeable lag on page load, and the FPS number stays high when the lag happens.

Reply Parent Score: 2