GTX 260 : 2000x1333 , video memory usage 1100 MB , pathtracing , 2.85 megasamples/sec , 0.7 FPS, render set to 1000 samples reached at 57 minutes, AA off, I tried with AA on and speed drops to 0.45 MS/sec
GTX 480 : 2000x1333 , video memory usage 424 MB , pathtracing , 11.09 megasamples/sec , 4.16 FPS, render set to 1000 samples reached at 4.3 minutes, AA off / on same speed

GTX 480 : 3000x2000 , video memory usage 633.9 MB , pathtracing , 12.91 megasamples/sec , 2.15 FPS. render set to 1000 samples reached at 8.55 minutes, AA off /on no speed change :O
Picture cleared after 5 minutes on GTX480 vs 1 hour almost in GTX260.
Good is't it? Higher the resolution better the samples, but what with the memory consumption? GTX 260 is far worst and it takes 1100 mb Vram. Not to mention the visual differences at same samples. I recheck all the textures to make sure that all materials have loaded their textures. No warning in the time Octane loaded the scene and all the texture was in place.. My scene contains a lots of glossy materials, bump, normal maps, specular materials ( lots of bottles and cups ) , lots of metal parts and geometry details and some clip maps.. It's and exterior and interior ( camera looks at the interior trough glass ).
Just a though, after this test I think we should made different benchmark scene to show the power of octane with fermi, as for current we can't see it.
So, will we expect better memory management, more speed for fermi cards in 2.3?