Latency vs. Bandwidth - Game Performance

[Deus Excalibur]

Has anyone done extensive testing to see what impacts games the greatest? This is with an Athlon64 system in mind. For example if you have the same speed overclock(lets say 2600mhz) but one approach is:

9x289 1:1 ; 2.5-4-3-7 timings

and the other is:

10x260 5:6; 2-2-2-5 timings

Which approach would offer the best framerate for modern games?

[vlad]

Check out charlies thread.
But 1:1 is always better.
http://www.xtremesystems.org/forums/showthread.php?t=55318

[dnottis]

I've done alot of testing with this since I have to run 2T 1:1 or DDR486 1T at tigher timings. Everything from 3dmark to q3, ut 2004, aquamark 3, doom III and HL2 showed faster performance with the memory at 1:1 2T than using the divider. I am going to eventually put together an article about this.

Here is some testing I did awhile back with a different CPU.

http://www.3dxtreme.net/other/CPUS/3000%20CBBFD%200441UPMW/1Tvs2T.jpg