I am also not convinced on the accuracy of your testing. From the fiddler trace you had it looked like the .png images came down with the original post (non-ajax pages) and cached at that point. The diffence in the page size of the AJAX vs NON-AJAX pages would have been down to the difference in the size of the markup of the two pages. I.e. the full markup for the NON-AJAX page and the partial page markup for the AJAX page. From the example it isn't immedialtely clear whether the images are being posted each time for the NON-AJAX part of the load test.