Originally posted by: Just learning
I've seen one benchmark (in the form of a simple graph) where a 878 Mhz 1440 shader HD5850 actually did better than a 1600 shader @ 850 MHz HD5870......but the HD5850 had its memory @ ~1400 Mhz.
This tells me that in that testing scenario memory bandwidth was the bottleneck....not GPU power.
Let me see if I can find the graph. I am sure it was probably 2560x1600 with heavy AA/AF for that to happen though.
From what I've read in this thread: http://www.xtremesystems.org/f...howthread.php?t=235181, it seems that neither memory bandwidth nor core speed are exclusively the bottleneck on the 58xx so far.
I think it would be errant to suspect that AMD would have so badly miscalculated the bottleneck of the card to the point where the thing was only memory bandwidth-starved -- they know that consumers for high-end graphics cards are going to see the benchmarks in modern games...