Scripps Employees Called 'Hackers' For Exposing Massive Security Flaw
Suspected PayPal Hackers Seek Deal To Stay Out Of Prison
3-way hardware-accelerated browser shoot-out: Chrome on top, IE9 just behind and Firefox brings up the rear (video)
After yesterday's announcement that Chrome 7 is now hardware accelerated, I instantly wanted to get the major browsers back into the ring for another screencasted deathmatch. Back when I did the 4-way speed test, only Firefox and Internet Explorer 9 featured hardware acceleration, and as a result Opera and Chrome were many orders of magnitude slower. If you watch the video, however, you'll see that's definitely no longer the case: Chrome is now the fastest of the three major browsers.
That speed comes at a price! As I discuss in the video, Chrome might be faster, but it uses significantly more resources than either IE9 or Firefox 4. Firefox is some 30% slower, but at the same time seems to use less CPU and GPU time. IE9 seems to utilize the same amount of CPU time as Chrome, but a little less of the GPU -- and it's marginally slower as a result.
What I don't know is whether this is by design or not. You'll notice that the GPU never went far above 50% -- why, with three browsers open, does it not get closer to 100%? The resources are there to be used -- why not use them?! Likewise, my CPU is still only half-used even when all three browsers are drawing 1000 frantic fishes at the same time. If you're curious, the other IE9 test drive samples all provided similar results. I wanted to try Google's 'HTML5 rocks' sample gallery, but they intentionally used elements of CSS and HTML5 that aren't yet supported in Internet Explorer 9 or Firefox 4.
In the name of science, here's some more information about my process: the screen capture does slow down each browser by a few frames per second, but relatively the figures are still accurate. I saw a small deviation in FPS when I was only running one browser at a time (probably because my CPU has multiple cores). There are a few unknown variables too, like whether the CPU core usage is defined by the app, or by the operating system (but with Chrome using more resources than IE9, you can only assume that Windows isn't unfairly biasing its own-brand browser).
If you'd like to recreate my test, you'll need to enable hardware acceleration in Firefox 4 and Chrome -- IE9 has it turned on by default:
That speed comes at a price! As I discuss in the video, Chrome might be faster, but it uses significantly more resources than either IE9 or Firefox 4. Firefox is some 30% slower, but at the same time seems to use less CPU and GPU time. IE9 seems to utilize the same amount of CPU time as Chrome, but a little less of the GPU -- and it's marginally slower as a result.
What I don't know is whether this is by design or not. You'll notice that the GPU never went far above 50% -- why, with three browsers open, does it not get closer to 100%? The resources are there to be used -- why not use them?! Likewise, my CPU is still only half-used even when all three browsers are drawing 1000 frantic fishes at the same time. If you're curious, the other IE9 test drive samples all provided similar results. I wanted to try Google's 'HTML5 rocks' sample gallery, but they intentionally used elements of CSS and HTML5 that aren't yet supported in Internet Explorer 9 or Firefox 4.
In the name of science, here's some more information about my process: the screen capture does slow down each browser by a few frames per second, but relatively the figures are still accurate. I saw a small deviation in FPS when I was only running one browser at a time (probably because my CPU has multiple cores). There are a few unknown variables too, like whether the CPU core usage is defined by the app, or by the operating system (but with Chrome using more resources than IE9, you can only assume that Windows isn't unfairly biasing its own-brand browser).
If you'd like to recreate my test, you'll need to enable hardware acceleration in Firefox 4 and Chrome -- IE9 has it turned on by default:
- Firefox 4 -- grab a nightly build, navigate to about:config and add gfx.font_rendering.directwrite.enabled -- set it to 'true'
- Chrome 7 -- grab a nightly build and add the following flags to the shortcut before opening it: --enable-accelerated-compositing --enable-gpu-plugin --enable-gpu-rendering --enable-accelerated-2d-canvas
Comments
11
Subscribe to commentsDabblerAug 31st 2010 1:44PM
any ideas on how to enable this on Mac OS X? I would think there was some way to toggle settings in a profile xml file or something like that?
Sebastian AnthonyAug 31st 2010 2:48PM
There was some mention of Windows/Mac stuff in the Chromium blog post, but I didn't look closely (check for the 'ANGLE' thing).
Pallab DeAug 31st 2010 1:36PM
Excellent comparison. Now Opera, will you please hurry up with your h/w accelerated Vega.
tech1greekAug 31st 2010 2:10PM
I would like to point out that the latest public release of IE 8 had a JS error for me, the beta 6 version of Chrome had 1fps the whole time, and Opera's latest public release gave 25fps - 35fps.
Opera is such an underrated browser.
steveAug 31st 2010 2:26PM
The GPU acceleration is windows only in Chrome.
The Firefox GPU needs the about:config of mozilla.widget.render-mode=6, otherwise you are just accelerating part, right? Or have they changed the default recently?
All of them I think are using DirectX with the flag saying that they are not a game, thus the GPU only gives them limited resources. To get full GPU, mean to get near exclusive access to its power, and side by side tests would be impossible.
Sebastian AnthonyAug 31st 2010 2:50PM
Yes, there's that other config flag too -- but I changed it to 0, and I didn't notice a decrease in performance. Perhaps it affects the other non-fishy samples -- I didn't check.
re: GPU use -- even when I run one browser on its own, it still only uses 10-15% of the GPU. It's not that they are competing for GPU time -- they are simply limiting themselves to a certain amount, for some reason.
3tearAug 31st 2010 4:28PM
Go banana!
DamianAug 31st 2010 4:41PM
I remember a lot of early talk with Firefox acceleration and an nVidia engineer. The idea was that Firefox would not push the GPU out of it's low power state, that way Firefox would not increase your electricity bill or kill your laptops battery.
Sebastian AnthonyAug 31st 2010 4:42PM
That's a very interesting point, if it turns out to be the case. Got a link by any chance? :)
ParalitykSep 1st 2010 8:33AM
No Linux love, yet :(
karaSep 3rd 2010 3:01AM
the only really way to run this test is to run one brower at a time.