Unlinked memory clocking.

Associate
Joined
3 May 2006
Posts
614
Location
Leeds
I've been doing a little testing lately on the P5N-E SLI with both 6300's and 6600's and Crucial Anniversary.
Didn't notice anything odd until today, when playing with 1T vs 2T to get an idea of the performance boost. (Please note, these pi times are crap. The OS is't tweaked at all, the times are taken for stability and comparison purposes only)

Have a pi 32M run done on the 6300 @ 500* 7 with the ram running 1:1 at 4,4,4,8 2T. The time was 15M 47S.
Did a run using the 6600 @ 500*7, but this time with the ram running at 432 FSB 4,4,3,8 1T. Pi 32m time in this case was 16M 14S.

That didn't make much sense, so re-ran with a lower CPU speed on the 6600. Time was worse. Re-ran with 2T, again time was worse. Then upped the FSB to 1:1 with 4,4,4,8 2T timings (as for the 6300 run) and the Pi time was pretty much bang on the 6300's time. Now aside from the fact the 6600 should be quicker than the 6300 in the 32m calc (This is correct is it not? Or is the 6300 a better match for the 6600 in 32m?) is anyone else seeing a performance dip coming from running in unlinked mode?
One would assume 4,4,3,8 1T @ 432 FSB would be as quick as 4,4,4,8 2T @ 500 FSB, but it seems the unlinked dividers are killing the performance.
 
Back
Top Bottom