Yep, it can be some problem which OpenGL calls are used by the irrlicht 3d engine (is used as 3D unit) - dont worry about that if your real games work or OpenGLExtensionsviewer* works as benchmark!
Also dont forget to try GRASS (its OpenCL - compute on GPU) - at least non hal100 (NV 4xx) cards should work with OpenCL.
* since last major Update the OpenGLExtensionsviewer got more "real game testing" if you use the new +huge 3D Object King insted of known CUBE (much simplier than KING).
Problem of really valide OEV tests is the huge configurability (Multisamling, Anithros, one Object vs many objects (KING vs KINGS))
To compare GPUs on complete different computers/screens better use:
resolution up to 1280x1024 (1440x900), windowed (fullscreen off , that disables also VSYNC!!!) , Multisampling * 2, Anithros * 8 and single Objects (KING or CUBE, not KINGS or CUBES).
Many Objects KINGS, CUBES is even more real game test, BUT cpu usage is also much higher = faster cpu gives much faster FPS with same gpu.
If you want compare OpenGL driver speed diffs (by OS X updates) on YOUR computer , its best to reduce GPU load to measure the (most) little driver speed diffs:
windowed up to 800x 600, multisamling off, anitros off
We can also use that thread now collecting here also OpenGLExtensionsviewer tests.
Beside my hint to use the above (first) listed config for OGLEV, you can post also you fullscreen FPS (same other config, but highest res of your monitor).