I have a question about building a budget gaming system based on the famous E5200 budget processor at the lowest price possible.
I will be looking at gigabyte and ASrock mainboards (possibly from your articles on the 600$ budget gaming pc) with overclocking options in the BIOS, an E5200 CPU, dual channel DDR2 , radeon 4770 and 7200rpm SATA-II harddisk.
Now my questions revolves around the infamous FSB RAM ratio. Forums often dispute that 1:1 is the ideal FSB RAM ratio and that you can increase memory speed, but after quite some research and thinking about it I don't think there is any use to increase this ratio as extra memory speed cannot be used by the processor and as higher frequencies increases the latencies of the faster running memory.
The (theoretical) maximum bandwidth of the E5200 is limited by system bus (correct me if I am wrong), which is 6,4 GB / sec. (200 Million Cycles/second) * (4 since it's quad pumped) * (64 lines/interface) * (2 bits / line / Cycle) = 51,200 Mbit/s, or 6,400 MB/s, or 6,4 GB/s)
The bandwidth of DDR2 800 memory is 6,4 GB/s as well (400 Million Cycles/second) * (64 lines/interface) * (2 bits / line / Cycle) = 51,200 Mbit/s, or 6,400 MB/s, or 6,4 GB/s) and double channel memory gives you 12,8 GB/s theoretical bandwidth.
The difference being that lower clocked memory has significantly lower latencies and could run at approximately:
Memory timings: 3-3-3-7
I suspect that in theory, I would have the same bandwidth in both configurations but much lower latencies in the latter configuration.
Now the same theory applies if you start raising the FSB to 266, you could use dual channel DDR2 1066 memory and get a FSB RAM ratio of 1:2 and use dual channel DDR2 533 memory to get an 1:1 FSB RAM ratio wher ethe theoretical maximum bandwith of the E5200 at a 266 bus speed is 8,5 GB/s.
Again, DDR2 533 has much lower latencies than DDR2 1066 and this could be beneficial to the system.
This article has shown that higher memory speeds hardly improve system performance (h t t p : / / w w w .xbitlabs.com/articles/memory/display/core2duo-memory-guide.html) and that lower latencies have a much bigger effect on overclocked systems, which makes sense as the processor is the bottleneck of the system (if you exlude the graphics card)
Now my question is: I want to use the following configuration:
Processor: 3.75 Ghz
Multiplier: 12,5 (max)
Bus speed: 300
Rated FSB: 1200
FSB RAM Ratio: 1:1
Max cpu bandwidth: 9,6 GB/sec
am I better off buying expensive (dual channel) DDR2-1066 memory running at (approximately) 6-6-6-18 that will go up to 1200 and give me a theoretical maximum bandwith of 19,2 GB/sec,
Can I use more affordable DDR2-600 memory (or DDR2-533 that clocks to 600mhz) in dual channel configuration that will operate at timings around 3-3-3-7 giving me the 9,6 gb/sec bandwidth to achieve 100% effectiveness in communicating with the processor?
- Is dual channel memory really delivering twice the bandwidth in practise?
- Does the CPU use it's full theoretical bandwidth in practice?
- Are there any situations where a system benefits from memory running at a higher speed than the CPU?
Thanks for your help and sorry if I am being ignorant!
Its a budget gaming build, forget the numbers. You worry about something that is going to have smallest impact on its overall performance. Slower memory wont cripple it(im running my ddr800 as ddr600 with oc`d e5200 - no difference) but will pretty much limit your options for upgrading the cpu.