Apollo, Flash, WPF/E, WPF and Firefox Speed Tests
Alexey Gavrilov is the man. He ran a bunch of different speed tests on a bunch of Rich Internet Application technologies as well as browser and operating systems and ended up with some very cool results. He admits that the result sets are small, but he enlisted the help of other people to bring the number up to 73.
He ran a 2D animation benchmark between the browser and found that the Apollo Alpha was significantly faster than all of the other browsers. That kind of surprised me, but it does make for some interesting conjecture about how diversely Apollo can be used to deploy applications.
Hey then took on RIA platforms and found that WPF was the fastest at rendering 2D animation at 45 frames per second. Flex (as a cached swf using the cachedAsBitmap setting) clocked in second with 38.6. WPF/e ran it at 35.6875 frames per second while “stock” Flex seems to have lagged at bit at 25.6 frames per second.
The best part of the survey were some of Alexey’s observations. A couple stuck out for me:
5. Microsoft is serious about WPF/e — the traffic from Microsoft to these pages was probably 40x bigger than from Adobe and Macromedia combined (I know they are the same company now — but networks are still different). Good for them.
6. Apollo (Flex version) has extremely good performance and it’s faster and lighter than WPF. Unfortunately Adobe is not going to allow deeper integration with native applications, so there is no real competition between WPF and Apollo. Bad for Adobe.
I don’t really think the native application integration is a big deal, and Apollo does outperform WPF by a wide margin. I know native apps are a source of contention, but Apollo is a pre 1.0 product, so I think we just need to wait and see how it all plays out.
Alexey, you are awesome! Thanks for doing these.
[tags]Flex, WPF/E, WPF, Apollo, Firefox, Opera, Performance Test[/tags]
Posted in Rich Internet Applications
March 31st, 2007 at 8:48 am
[...] [via Ryan Stewart] Bookmark to: - Posted in Flash, Flex, Apollo by Sönke [...]
March 31st, 2007 at 1:35 pm
I haven’t evaluated the tests, and so can’t describe what they actually measure… sounds like just simple screen-drawing.
Browsers definitely do vary in how frequently they grant plugins access to the processor. I’m not surprised that even an alpha release of the Apollo runtime would offer 50-250% performance boosts… Apollo is actually a desktop application, not a document browser. (If he’s saying that Apollo alpha outperforms shipping Vista WPF, then that would shock me.)
The observation that Microsoft traffic into his first test was forty times as large as that from Adobe domains was interesting, because their staff is only about a dozen times as large as Adobe’s staff — higher per-capita rate. Their blog aggregator is also staff-only, so maybe they went to something like MXNA to see what customers were saying.
jd/adobe
March 31st, 2007 at 3:24 pm
Hey Ryan, thanks for posting this. Just a little clarification on what actually “Flex / cached” version was. It’s the version of the test, where cacheAsBitmap option is set for objects being animated (ie balls) unlike original version, where it all rendered in vectors. The reader pointed out in comments that turning this on can boost the performance of a Flex version (and it did for certain platforms) so I added it to the comparison.
March 31st, 2007 at 3:44 pm
Cool, thanks Alexey. I was kind of confused about that because I didn’t think the local cache of the SWF would make any difference. I’ll fix the post!
March 31st, 2007 at 6:34 pm
hmm…really interesting this comparison. but it’s good to remember that wpf/e it’s on his ctp version, and performance improvements will be made before the final product is arrived
April 1st, 2007 at 9:40 pm
I have a couple of questions about why we would even be benchmarking these things beyond asking the question are these platforms sufficiently capable for applications and deployment? It’s obvious that both of these platforms are going to have areas of advantage and disadvantages. A few FPS one way are another aren’t gonna tell the story of whether wpf (and wpfe) succeeds where flash/flex/apollo does or doesn’t.
Together these both represent significant movement forward for everyone. I’d love to see more people thinking about how we can use them together in our own mashups. I really feel they are quite different enough they won’t really eat into the other’s space, as they are different enough that they will find different purposes..
I am really not sure the benchmarking thing is very useful, as I wouldn’t use it as a business reason for recommending or not recommending an application be developed on the platform. I would really love to see more discussion about the business reasons for using this and when to make an app that works in the browser versus making it a desktop app..
Now if we were talking about hardware accelerated 3D in WPF versus something like PaperVision 3d maybe there would be more that would come to mind to discuss..
I guess I am hoping this spurs more discussion about issues like:
If I am making an internet app when should it just run in the browser, when should we use ajax/spry, when would an app be more appropriate outside the browser, and how to make the business case for each of these scenarios and why..
Anyway that’s just my thoughts. Thanks for hopefully starting off this dialog anyway.. Benchmarks are fun just the same, what’s that old saying “there are benchmarkes, lies, and damn benchmarks” can’t quite remember that quote..
April 2nd, 2007 at 12:36 am
Don, how useful are things like this? Maybe not that useful from a technical perspective, but they’re interesting, and it does generate a bit of conversation. I think you bring up a good point about having a discussion of in-browser vs. out-of-browser. Maybe I can put my thoughts into a ZDNet post.
April 2nd, 2007 at 5:14 am
i also think like that what ever you are thinking, what a considence, and i think it will be very helpfull for us and also that is a good platform for us to discuss.