HTML5 Performance Tested Against Flash
ReadWriteWeb distills some results from a Streaming Media post about performance between Flash Player 10 and 10.1 and HTML5.
In analyzing the results of the tests, Ozer determined that the key to better Flash performance was dependent upon whether or not it could access hardware acceleration. This feature, launched in Flash 10.1, allows the plugin to use the graphics processing unit (GPU) on some computers to decode video. Depending on the video card and drivers, (NVIDIA, AMD/ATI and Intel offer products that support this), the video decoding process in Flash 10.1 can now work for all video playback, not just full-screen playback as was available in Flash 10.0.
The streaming media post has some good information about the method used to collect data and it seems like a pretty robust test. He used different machines and collected 29 data points for each test. So this isn’t some isolated test case.
But what I was most happy with was the performance of 10.1 across the board. The Flash Player team has been working their butts off to make Flash Player 10.1 scream and it looks like it’s paying off. The results show huge drops on Windows and good incremental improvements on the Mac. Mac isn’t on par with Windows but that’s because our engineers don’t have access to the same APIs that they do on Windows to get that boost from hardware acceleration.
I think Flash Player 10.1 is going to be a great release and I’m excited to see a leaner, meaner Flash Player get penetration.
Posted in Adobe, Flash Player
March 10th, 2010 at 1:59 pm
To be fair, it’s not a case of access to APIs, so much as existence of equivalent APIs.
March 10th, 2010 at 4:39 pm
@JulesLt, without seeing the code for Quicktime in OSX, is it possible to make the distinction? Quicktime has access to hardware H.64 decoding – and OSX desktop apps like Chrome consequently have access to it through the Quicktime API – so is Quicktime’s access based all on custom, low-level code or is there a private API in OSX that it’s accessing? I guess it also depends on your defintion of API but I would imagine that there’s some sort of interface that Quicktime is using that might easily become a ‘public’ API.
March 13th, 2010 at 4:53 pm
I don’t know about that, this performance increase is across the board. And we obviously had to make Linux better to work on Android devices.
=Ryan
ryan@adobe.com