With 333 FBS 1:1 DDR2-667 or DDR2-800 using divider

gstem

Distinguished
Mar 12, 2007
8
0
18,510
I'm going to over clock my e6700 by raising FSB to 333 (1333). Would it be best to use DDR2-667 or DDR2-800 each with CAS 4-4-4-12. The first would, of course, be in a 1:1 ratio, yet would the 800 be a large enough frequency increase to overcome the loss of performance from using a divider?
 

graysky

Distinguished
Jan 22, 2006
546
0
18,980
1:1 = no divider. The DDR2-800 will allow you to hit a FSB of at least 400 MHz, but @ 333 MHz, both the chips will be the same given the same 4-4-4-12 timings.
 
I think what the OP is asking is that if he only wants to go to 333 FSB, should he leave the RAM at 1:1 for DDR2-667, or use the next ratio and get around DDR2-800 RAM speed from the 333 FSB.

I'm also very curious about this since I run 333MHz full time. According to this madshrimps article, there is a penalty for running at 667 (5:4) speed so it is actually better to run at 533 (1:1) until you hit 800 speed (3:2) when the increased frequecy overtakes the penalty of running asynchronously. This is with a 266 FSB, but at 333 FSB, the 1:1 RAM speed is now 667 and the 5:4 ratio is now ~833 speed. Would the asynchronous 833 RAM speed be too big a hit compared to sychronous 667 at FSB 333MHz? Is my math right with the ratios?
 

gstem

Distinguished
Mar 12, 2007
8
0
18,510
Thank you rwpritchett for better articulating my question. Our hypotheses is identical. I hope there is someone out there who has tried ddr2-667 and 800 over a 333 FSB to prove the analogy to existing articles over 266 FSB (which compared 533 to 667).
 
I played around with the 3DMark benchmark when I first set up my system. I too found that the frequency did very little to improve performance. Timings, however, showed a much bigger improvement. For example, my system scored higher at DDR2-533 @ 3-3-3-8 than at DDR2-800 4-4-4-12.
 

little_scrapper

Distinguished
Aug 2, 2006
621
0
18,980
just this last week, new build, I have run 3DMark05 with my E4300 at 9 x 333 using both 667 and 833 memory. Both running 4-4-4-12. with my memory multiplier at 2 for 1:1 DDR667 I scored 6615 marks. With the memory multiplier set at 2.5 for DDR833 I got 6650marks. So I saw almost a 1% increase!!! WOOHOO!!! Glad I bough the cheap Patriot Extreme ram.

Basically since the memory is double data rate, when its running at DDR667 its effectively double clocking for 1333Mhz. Which is equal to the CPU when the FSB is at 333. So that any memory running faster, ie DDR800 is going to have to wait for the CPU since DDR800 is effectively running at 1600MHz.

I saw about a 1% increase going from 667 to 833 keeping the same timings. Now if I went to DDR900 and dropped from 4-4-4-12 to 5-5-5-15 I would probably start to see a loss due to the slower timings.