Faster memory means I should do a higher ratio?

[Space_Coyote]

I just upgraded from 2x1GB Crucial Ballistix to 2x2GB Geil. As usual, I’m running my E4300 at 3.0 GHz (333 FSB). Before, I was running at 1:1 FSB to RAM ratio, which never even came close to maxing the speed of the Ballistix. Should I be running the Geil at a higher ratio to take advantage of the faster speed?

ATM I have the Geil at 1:1 ratio, 2.0v and tightened timings to 4-4-4-12, with 9 hours and counting Orthos blend test stable.

 

[kaiweiler]

I like running my RAM at 1:1 ratio, but if I were you I would try something like 375*8 or 430*7 or something and run some benches to see if any performance is gained.
But overall as far as I know it is best to keep RAM at 1:1.

 

[Space_Coyote]

I'm still new at this, so forgive me if this is a stupid question. Forgetting the fact that the memory is double the size, what advantage is there in having DDR 1000 memory over DDR 800, if I'm only running them at 666?

 

[kaiweiler]

only that when you run RAM below it's rated speed you can often tighten the timings up.
for example some DDR2 800 with 5-5-5-15 timings can often run with 4-4-4-12 timings when at 667 speeds as opposed to the rated 800Mhz.

 

[BillParrish]

I'm still new at this, so forgive me if this is a stupid question. Forgetting the fact that the memory is double the size, what advantage is there in having DDR 1000 memory over DDR 800, if I'm only running them at 666?

Very little to none. Waste of money actually.

When the numbers work out right, I also like running 1:1 memory ratios as the system is designed that way. Running a mulitplier requires wait states and buffers to keep timing correct. However Intel has done a lot with memory controllers to help reduce that penality and while there is some performance penality the potential increase in performance due to the increased bandwidth far exceeds it.

If you were only a little off, I would say benchmark and see, but you are way way underclocking that memory and its a waste. Find a good mulitplier that gets you close and go with it.

I also agree with kaiweiler, if you board will remain stable another good option is to drop the cpu mulitplier and raise the FSB to give the same or better CPU speed but to run the memory faster at 1:1 or to find a multiplier that is more stable (testing and benchmarking and keeping notes would be critical for finding the "sweet spot". )

So a combination of the two methods will probably be best depending on how well your board handles increased FSB.