Completely Unscientific Test of Chrome 2.0 Beta with Bubblemark
So, I had to rush out and get the Google Chrome 2.0 Beta and try it out.
I was already running Chrome 1.0, but the install for Chrome 2.0 installed with no hitch, asked me to restart the browser, and then I see that all my settings/bookmarks migrated forward with no problems. After playing around with some the new features I had to find out how Chrome's DHTML Rendering performance stood up to Flash/Silverlight/JavaFX.
Bubblemark is a standard benchmark for comparing these different technologies. It simply tries to render a certain number of images (or vectors in some cases) and animate them. I fired it up and collected some numbers. Here are the results of my _totally_ unscientific test:
Suprisingly to me, the LWJGL libraries were the fastest of them all. I hadn't really even considered using it for RIA, and I don't know if there are any component libraries out there that might be useful for creating applications with.
Also interesting is that the Swing (Optimized) demo actually scaled much better than any of the other options. JavaFX appears to needs more work as it had a nice strong start, but couldn't really scale up to the 128 ball test.
Another interesting observation is that DHTML actually scaled better at the higher end than two of the "hotter" RIA technologies out there today (Silverlight and Flex).
The final surprise for me was that the cacheAsBitmap optimized Flex benchmark couldn't match pure DHTML performance in either Chrome or IE.
Obviously, benchmarks aren't really a total picture of things. First off, no one is going to develop and RIA that animates a bunch of bubbles. Lots of RIA will probably simply sit there, for the most part, waiting for user input. You also have to take into consideration things like development time, ease of use/maintainability/etc. But it does seem to open my eyes a bit more about how the playing field for RIA stacks up right now in terms of performance.
I was already running Chrome 1.0, but the install for Chrome 2.0 installed with no hitch, asked me to restart the browser, and then I see that all my settings/bookmarks migrated forward with no problems. After playing around with some the new features I had to find out how Chrome's DHTML Rendering performance stood up to Flash/Silverlight/JavaFX.
Bubblemark is a standard benchmark for comparing these different technologies. It simply tries to render a certain number of images (or vectors in some cases) and animate them. I fired it up and collected some numbers. Here are the results of my _totally_ unscientific test:
Browser + Tech | Bubble Count | FPS |
---|---|---|
Silverlight 2.0 (CLR) on Chrome 2.0 | 16 | 283 |
128 | 44 | |
Flash (Flex) with cacheAsBitmap on Chrome 2.0 | 16 | 115 |
128 | 46 | |
Java (Swing, optimized) on Chrome 2.0 | 16 | 200.0 |
128 | 199.7 | |
Java (LWJGL) on Chrome 2.0 | 16 | 1435 |
128 | 658 | |
JavaFX 1.0 (at Osvaldo Doederlein blog) on Chrome 2.0 | 16 | 315 |
128 | 21 | |
Pure DHTML on Chrome 2.0 | 16 | 176 |
128 | 85 | |
Pure DHTML on IE 7.0.6001.18000 | 16 | 64 |
128 | 85 |
Suprisingly to me, the LWJGL libraries were the fastest of them all. I hadn't really even considered using it for RIA, and I don't know if there are any component libraries out there that might be useful for creating applications with.
Also interesting is that the Swing (Optimized) demo actually scaled much better than any of the other options. JavaFX appears to needs more work as it had a nice strong start, but couldn't really scale up to the 128 ball test.
Another interesting observation is that DHTML actually scaled better at the higher end than two of the "hotter" RIA technologies out there today (Silverlight and Flex).
The final surprise for me was that the cacheAsBitmap optimized Flex benchmark couldn't match pure DHTML performance in either Chrome or IE.
Obviously, benchmarks aren't really a total picture of things. First off, no one is going to develop and RIA that animates a bunch of bubbles. Lots of RIA will probably simply sit there, for the most part, waiting for user input. You also have to take into consideration things like development time, ease of use/maintainability/etc. But it does seem to open my eyes a bit more about how the playing field for RIA stacks up right now in terms of performance.
Comments