Well isnt some of this down to the GPU being the absolute bottleneck- especially in single card?
After that, id put some of the scaling down to the 3DM score formulae and coding
The score improvements would be less by moving from 3-4 cards as the balance of power moves closer to equilibrium...maybe even tip the bottleneck back to the CPU?
Having said that, results like these make me quite happy about how ATI have implemented their multi-GPU solution. Good engineering well implemented.....finally!




Reply With Quote
Bookmarks