^ that is the trick.....
If Intel, use 800mhz.
IF AMD, use 1066.
Intel won't benefit much from faster memory, AMD has the integrated memory controller on chip, and will show improvement with faster memory.
Maziar, I have 4 1066mhz CL5 ram chips in my machine so heres my experience.
2 chips you can hit speeds well above 800mhz without bring up the fsb to 400mhz (1600mhz qdr). 4 chips makes going over 800mhz difficult. Sure the ram will be set to higher than 800mhz but the bandwidth will still be limited. Its a northbridge limit. Secondly I run my chips at 875mhz CL4 which is way faster than 1066mhz CL5.
Hell on my custom 3DMark06 runs (1920x1080 8xAA 16xAF) it only adds about 500 points from changing the ram settings. :whistles:
Me E8300 is at 4Ghz, meaning I have to run my RAM at DDR 2 1000 to match it. My RAM is 800 stock. If you plan to run it 1:1 and could get up to 1000 you might as well get RAM that is already there, that way when you increase the voltage you can losen the timings, while in order for me to hit 1000 I had to up the timings to 5-5-5-15 from 4-4-4-12 (hit that wall at around 925).
The AM2+ will see a benifit with 1066 (ONLY WITH PHENOM, OR TRI-CORE!). The Intel will get more of a benifit from RAM with good (4.4.4.*) timings. The older AM2 (ie the 5000+ BE, or any X2 proc) will not, I say again will NOT! be able to use higher than DDR2 800 (memory restrictions on older procs). The reason for this is the AMD chips have an onboard memory controller (the intel does not.............YET!)(wait for Nehalem for the onboard Mem controller)