AnsweredAssumed Answered

glGenerateMipmap very slow on Radeon HD cards

Question asked by bugbuster on Apr 15, 2013
Latest reply on Apr 19, 2013 by bugbuster



I'm developing a 3d game engine based on openGL. Recently I've made some performance test between 2 different systems:


AMD 64 dual core with Geforce GTS 250 versus Intel i7 quad core with Radeon HD 5850.


Incredibly the rendering is faster on the first system. I've used GPU time counters (timer query extension) and CPU counters and I've found that RadeonHD 5850 is faster than Geforce GTS 250 (GPU counters give small numbers, = minor time spent, in favour of the RADEON (of course), I7 is faster than AMD 64 (when doing operation on CPU i7 is of course faster) BUT when I do a call to OpenGL command, Catalyst drivers results slower than NVIDIA (very very slower), so that the first system run faster!!!!!

In particular I've made a test on glGenerateMipmap for a 128x128x6 RGBA cubemap and following are the results:

- on Geforce GTS 250 the call time is 0.3 ms

- on RadeonHD 5850 the call time is 3 ms!!!!!!!!!!!!!!!!!!