Originally posted by: ajmiles
So why don't you just run Sandra as your main app ? Seriously the real world difference is very very small, the reason is that the memory is rarely working at full bandwidth. And for the non-believers, here are some hard measurements, compare for yourself the performance at 12x200 to 9x266. Read it and weep.
There is no proof on those two links at all user1234.
The sandra memory tests show a sizeable increase is what you expect, however there is no application other than sandra that uses a lot of RAM to it's full potential. Quake 3, RtCW at 1024x768?? Where is Half Life 2 and Doom 3 at 1600x1200 with 4xFSAA and AF? That is where the advantage will be seen.
As for their being no difference with the SuperPI tests.... of course not, the memory that program uses will probably fit in the L2 cache anyway.
it's laughable that you think that cpu memory access pattern will change when you run a game with or without aa/af, which are graphical transformation algorithm performed solely by the graphics hardware and does not even entail additional communication with the CPU. Also for higher resolutions, the memory access patterns of the cpu will not change much.