FSB or Multiplier

[polobunny]

Originally Posted by Badluck Inc

Sorry was a typo its running at 1066 , CPU-Z say DRAM freq is at 533 . Should my FSB:DRAM be 1:1 like in the guide or does that really matter?

If you can manage to get a 1:1 ratio you will get a nice boost on memory performance. It's not always possible with the limitations of your CPU and motherboard though.
I was talking about 517x7 earlier, which would make your quad run at 3.6GHz and your RAM at 1034MHz on 1:1 ratio.
You would need to check first if it's possible to run a 517 fsb with your current configuration (should be alright, though it's a high FSB you have a nice chipset), then if it's worth it AKA if there's a performance boost of running your RAM at 1034MHz 1:1 compared to 1066MHz whatever divider.
I like to think the performance of running your RAM at 1034MHz 1:1 would overweight 1066MHz at your current divider.

[Badluck Inc]

Thanks for all the help with this . Right now i have it set up at 400x8 which gives me 3.2 gigs and ram at 800 which gives me the 1:1 . I know i can go alot higher but i want to keep things low atm .

[polobunny]

Originally Posted by Badluck Inc

Thanks for all the help with this . Right now i have it set up at 400x8 which gives me 3.2 gigs and ram at 800 which gives me the 1:1 . I know i can go alot higher but i want to keep things low atm .

To test for memory bandwidth you should check using Everest or SiSoft Sandra. There's a free version of Everest that's more than enough to check that out, and there's a trial for SiSoft Sandra.

[gvblake22]

Those synthetic memory bandwidth and latency benchmarks are good for finding very specific performance differences between different RAM settings (like what you are testing out here), but just remember that synthetic benchmarks are not typical of real world performance. To put it another way, there may be a significant difference between two different RAM settings according to EVEREST benchmark results, but you may never actually notice a difference when just using your computer for what you usually do. It's just to know that you are running with settings that are the fastest and not something that could be slower than another setting you could be running because you didn't test it to see if it was "theoretically" faster (according to the synthetic benchmark). Just something to keep in mind when running benchmarks from EVEREST or SANDRA.

[polobunny]

Originally Posted by gvblake22

Those synthetic memory bandwidth and latency benchmarks are good for finding very specific performance differences between different RAM settings (like what you are testing out here), but just remember that synthetic benchmarks are not typical of real world performance. To put it another way, there may be a significant difference between two different RAM settings according to EVEREST benchmark results, but you may never actually notice a difference when just using your computer for what you usually do. It's just to know that you are running with settings that are the fastest and not something that could be slower than another setting you could be running because you didn't test it to see if it was "theoretically" faster (according to the synthetic benchmark). Just something to keep in mind when running benchmarks from EVEREST or SANDRA.

But then again, if it proves to be faster in synthetic benchmarks, it will be faster in real world applications as long as the benchmark is reliable. And that can't hurt. :P

[gvblake22]

Originally Posted by polobunny

But then again, if it proves to be faster in synthetic benchmarks, it will be faster in real world applications as long as the benchmark is reliable. And that can't hurt. :P

Yes, that's what I was trying to say. I just wanted to add that synthetic benchmark performance is often misleading when it comes to something you will actually notice when using the computer. Sometimes it will, sometimes it won't.

[Boy'nBlack]

I ran those marks Sunday afternoon and walked away a bit baffled at the results, so I hesitated to post them yet. I ran 350x8 and 400x7 and they're identical under a quad, while giving the nod to the 400x7 by 5% on a dual.

Since they were tests on different platforms, I really need to try the dual in this computer and have another go at it. Or perhaps not use synthetic runs...not sure.

[polobunny]

Originally Posted by Boy'nBlack

I ran those marks Sunday afternoon and walked away a bit baffled at the results, so I hesitated to post them yet. I ran 350x8 and 400x7 and they're identical under a quad, while giving the nod to the 400x7 by 5% on a dual.

Since they were tests on different platforms, I really need to try the dual in this computer and have another go at it. Or perhaps not use synthetic runs...not sure.

FSB or multi doesn't matter much at some point, although a higher FSB usually gives better results (but strains more the chipset, I believe).