Product Spotlight
(3 votes)

Silverlight rendering performance in different browsers

4 comments   /   posted by Denislav Savkov on Sep 30, 2008

Introduction

Recently we started wondering if there is a noticeable difference in the performance of Silverlight in different browsers.  Measuring performance reliably is not a trivial task.  One thing is that performance have many aspects like rendering graphics and manipulating large data. Obviously to rate overall performance you need to balance the weight of individual aspect ratings.  Finding the appropriate balance depends on the purpose of the rating.  Since Silverlight's primary task is to create RIA  we decided to test first the graphics rendering performance.

Benchmark

There are a few popular benchmarks that compare Flash vs. Silverlight and other technologies like Bubblemark and GUIMark. With bubblemark results were varying  from one run to another. So to reduce the number of tests we chose GUIMark because the results were more similar. We made a few changes to it though. We added more gradients and transparency and removed part of the text. Although the test may be not entirely reliable for measuring maximum performance it souits our purpose because our goal is to compare performance under the same scenario.

Test procedure

Our test machine was desktop Athlon 64 3000+ with 1GB of RAM and Windows XP SP3 installed.  The browser versions that we tested are: Internet Explorer 7; Internet explorer 8 Beta 2 - 8.0.6001.18241; Firefox 2 - 2.0.17; Firefox 3 - 3.0.2;Safari - 3.1.2. We ran the same benchmark three times on each browser for 10 minutes. In the end we took the average frame rate. The current version of Silverlight that we used is 2.0.30253.6. labeled Beta 2.

Results

The difference in performance is not that big. Actually the difference is so small that you won't notice it with naked eye. Having in mind that we test the same plug-in this result is not surprising. Internet explorer 8 Beta 2 and Firefox 2 are slightly faster than IE7, Safari and Firefox 3 but you may choose anyone to be your without concerning Silverlight performance. Performance difference is much bigger if you start some other application than if you change the browser.

  Firefox 3 Internet Explorer 7 Safari Firefox 2 Internet Explorer 8 Beta 2
fps 22,11 22,17 22,23 22,44 22,5

It is curious that when you instantiate second browser with the same test running the performance drops at half the framerate and is equal for both instances.

Conclusion

As you can see the difference in the performance is not that big. Actually the difference is so small that you won't notice it with naked eye. Having in mind that we test the same plug-in this result is not surprising. Internet explorer 8 Beta 2 and Firefox 2 are slightly faster than IE7, Safari and Firefox 3 but you may choose anyone without concerning Silverlight performance. Performance difference is much bigger if you start some other application than if you change the browser.

Finally

We will continue with more performance tests of Silverlight that is why we are interested in testing scenarios that you are interested in, so please leave a comment and we will include them in our future tests.

Reference

http://www.craftymind.com/guimark/

http://www.bubblemark.com/

Filed under: Misc


Comments

Comments RSS RSS
  • RE: Silverlight rendering performance in different browsers  

    posted by Michael on Oct 06, 2008 20:10

    Your graph looks like it has IE6, I think it should be IE7

  • RE: Silverlight rendering performance in different browsers  

    posted by iiordanov on Oct 06, 2008 23:39

    Hi Michael,

    You are absolutely right! I will update the graph today.

    Thanks for your comment on this.

  • RE: Silverlight rendering performance in different browsers  

    posted by Sean on Oct 16, 2008 10:57

    Any chance we can see what the modified guimark testcase looked like?

  • RE: Silverlight rendering performance in different browsers  

    posted by iiordanov on Oct 17, 2008 07:53

    Hi Sean (Christmann?),

    Yes, we will send you the modified project, but in general the changes are minor. Btw we sent you an e-mail with a request for permission to publish the modified source code in this article, but we did not get an answer...

Add Comment