I have received a number of questions from users of my software about low performance on ATI HW I am not using any ATI extensions but i use ext_draw_array_range and compiled vertex arrays so I couldn’t really understand their issue.
I just got a ATI FireGL X1 in my post to be able to add VAO etc. I started to test a simple software with just plain tri strips and triangles and found a horrible low performance compared to my GForce 3 and GForce 4
Ok. Im an not using any ATI extensions but why is it so MUCH lower performance. It this driver related. I get about 1 M TRI per second ??
I made some tweaking. Removed the compiled vertex array extension and I get 2.5 M Tri per second. Strange that the extension makes it run slower ?? In my locked arrays I use all indices.
What performance do you get on a radeon 9700 with the latest drivers using only gzDrawArrays on tristrips ??
Do the Radeon 97000 have later drivers than the FGL 9700 ?
Performance on “basic” OpenGL is worse on ATI cards than on NVidia’s. I found 50% to not be unusual.
ATI drivers are much more dependant on the CPU power than NVidia’s. I’m not surprized that you get so low performance if you “only” have 800 Mhz.
ATI drivers/cards are very dependant on the number of API calls you make; basically, more than 1000 or 2000 glDrawElements (or equivalent) calls per frame will kill your performance, whereas you can get 10x times more on NVidia’s without seing a slowdown.
Performance is best with VAO, but strangely i found that they optimized their driver quite a lot for immediate mode calls. Try using IM and see if it’s faster than VA/CVAs…