I have read up on RAM dividers but cannot find a clear answer. I OC'ed my e6600 to 3.0 so FSB = 333. So which is better 1:1 where my ddr2-800 runs at 667 or set the RAM divider to 5:6 so my RAM runs at 800? I assume the latter is better? Is there a reason to do 1:1 and have your RAM run slower than its rated? Or is having your RAM run as fast as possible better?

Divider = RAM FSB ratio to ( CPU FSB

Lets say you've RAM that will work DDR800 = FSB400 on 5-5-5-15 timings.

You've overclocked your CPU/MOBO FSB to 333MHz, thus making RAMs run 666/667MHz on 1:1 divider.

When you change to divider to 5:6 your RAM runs 667 * 5/6 = 556MHz

There's hardly idea running your RAM that slow, you'd want to set them 7:6 divider or something that would make your RAM run 667 * 7/6 = 778MHz which is closer to 800MHz.

Thing is, when you do some "serious" overclocking you might set your FSB to 500MHz. For DDR memory this would mean 1000MHz speed. Unless your RAM can run that fast (you can make em run faster, by lowering timings), you might need to use a divider, 5:6 in this case.

This way CPU/MOBO = FSB 500 = 500MHz
And RAMs = 500 * 5/6 = ~416MHz *2 (DDR -> Dual Data Rate) = ~833MHz. This is such a minor OC for the mems (under 5%) that they probably can take that.

Here let the pictures do the talking. Pay attention to the "read speeds" and "latency".

Err. I've hard time believing you're getting proper info from that Lavalyst product. Get cpu-z instead. Idiotic info from that one, though it's correct.

Your normal FSB for memories is 400MHz. Which would be 1:1 (8:8). You're running your mems at 600MHz = 12:8 compared to normal clocks. 600:400 MHz. That's why you're seeing 12:8. Naturally latency is less when your mems run 1200MHz instead of 800MHz.

I have CPU-Z but it doesn't have a memory bechmark and thats what I was reporting, Sandra reports memory & latency are pretty much the same higher read & lower latency. CPU-Z shows a FSBRAM is 2:3 ratio.

I got the 5:6 number from CPU-Z which according to the first poster is incorrect becuase my RAM is running at 800Mhz so I am not sure what the correct divider is, but CPU-Z seems to be showing the incorrect ones.

So it seems once your OC'ed your CPU and your happy with it, then you should choose a divider to run the RAM at least asclose to its stock speed?

Well with DDR2/DDR3 and ram in general faster seems better as long as the clock speeds stays high enough to where the latencies are still low. The faster I run my ram the lower the latency & higher the read/write speeds are.

1:1 is still best and more stable. Just stick with 1:1 and instead of freq. try to lower the latencies as much as possible.

People have been saying that to me forever and yet I show proof that it isn't so and you still say, "that 1:1 is still the best and its more stable." I have shown in the pictures above that higher speeds are better. So can you show me if what you are saying is true because I would like to know one way or the other. Thank you.

You dont have to take advice you can do whatever you want. Lower latency's to that extent isnt enough to make up for that decrease in frequency, but its still best to go 1:1.

I too ram memory benchmarks using Sandra, as you would expect, when the RAM is running at 800Mhz it is much faster than 1:1 and 667Mhz. I just wanted to make sure I was not aware of some special case that makes it worth running 1:1 better than some other divider. Something along the lines of since the FSB:RAM are synched that overall system peformance is better, etc. As this thread shows, and from reading posts online I cannot find a clear answer. If it's a stability thing, then I can understand 1:1, but if your system is stable at some other divider I do not see how having the RAM run at its stock speed could be a bad thing. In most cases faster is better, no?

If 1:1 is better then what the heck is the PC world doing creating something called DDR3 @1800MHz, shame on them.

They're doing it to keep discussions like this going and to keep the overclockers happy.

*goes back to turning a D805 and 8800GTS into liquid metal*

Take a look at the following guide;
http://www.overclock.net/intel-cpu [...] cking.html

It suggests that 1:1 may not be the best with some explanations. Seems the controversy continues.

Heh Iv done a lot of reading onto this memory stuff as of late, it seems that 1;1 is not "best" but rather, the memory ability relies on the FSB for bandwidth, so say you run memory at fsb memory 1:2 it wont have enough bandwidth. And running 1:1 kinda garentees that.