Hello Adam, Thank you for the explaination. I will check for the overhead in setting hardware acceleration. If the overhead cant be reduced, I will try to set software rendering for the small operations and check if "Xfbdev with exa" still drops performance compared to only "Xfbdev" for gtkperf. I understand the gtkperf is not the right application to check for hw acceleration, however I atleast need to get the performance numbers equal to no hw acceleration (no exa). <div class="gmail_quote">On Mon, Aug 9, 2010 at 11:44 PM, Adam Jackson <<a href="mailto:ajax@nwnk.net">ajax@nwnk.net</a>> wrote: <blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> That you see a performance drop for small operations in x11perf indicates that your overhead for setting up accelerated operations is dominating rendering time. Fix that first. After that remember that most of the GtkDrawingArea tests use core rendering primitives that modern applications generally don't use. Also: there was no reason to direct this only to me. Send future correspondence to the list so that others can learn from it, and so that if (heaven forfend) I'm wrong about something that someone will correct me. - ajax </blockquote></div>