I've been wondering about the FSB:RAM ration. How does it effect the overall performance anyway? I don't need to know it in the details, just a simple explanation would be more than enough. 1:1 vs X:X?
An example would be a Q6600 running with 800Mhz RAMs. The FSB:RAM ration would be 2:3.
Does that mean that the RAM is running at lower clockspeed than it can handle? (400 x 2/3 = FSB speed of 266Mhz. 266Mhz x 2 =~533Mhz at clockspeed and 1066 at dual channel).
Would there be any different in performance if I were to replace the 800mhz RAM with a pair of PC-4300 533Mhz native RAM modules? (Don't mind about OC please, it wouldn't be any if I use the PC-4300 that I know)
The second question would be about the 800Mhz RAM clockspeed.
I'm planning to overclock my FSB to 400Mhz. My E6850 will get around 3.6GHz and stays there. The RAM clockspeed will be 400 x 2 per stick and in dual channel it will be 800 x 2 = 1600MHz.
1600Mhz, can this be considered as a strain on the RAM module? How 'safe and stable' can it get at a clockspeed like this? Enough for running 24/7 with it?
If your ram is running at 400MHz x 2 and you have DDR2 800, its not being stained at all its running at stock speeds. As for your FSB-Dram 2:3 means you ram is running faster than the CPU FSB. This can be a good thing as your processer has more memory bandwidth. When I get my ram fan to cool my dimms I'll be running my ram at 1200MHz 5-5-5-15 @2.2v which gives me a 2:3 FSB-Dram ratio.
Means you're will be pushing your FSB all the way up to 600MHz?
Isn't it better with 1:1 ration? The FSB and RAM would be in perfect sync then?
To your first question No the ram will be at 600MHz x2 dimms, the CPU FSB will still be at 400MHz. Two the second question you might think so but with all the DDR3 speeds you have to ask yourself why are companies coming out with ram that does 1600-2000MHz if 1:1 is better. Some will say to get your money and keep overclockers happy, but is that really why? You may not notice the extra bandwidth in all applications, but look at these benchmarks and pay attention to "Read & latency>>
Yes, I notice a very good improvement in both read and latency.
But isn't this kind of benchmark which test RAM and CPU separately? Or does it show the test result when CPU and RAM working together?
I'm still a little bit confused about the ration.
Join the rest of us. People have been saying to me that 1:1 is better, but then I run a simple benchmark that proves otherwise. On top of that now we have DDR3 with speeds over 2000MHz to make things even more confusing.