FSB DRAM Ratio Question

mikeynavy1976

Distinguished
Feb 14, 2007
452
0
18,810
16
I have a question on the FSB:DRAM ratio issue I keep hearing so much about, and what I should do to optimize my system. Right now I have:

Motherboard: Gigabyte EP45-DS3R
CPU: Intel Q9450 (OC'd to 3.2GHz via 400 x 8)
Memory: Corsair Dominator PC8500

I just changed my FSB:DRAM ratio to make it 1:1 per so many recommendations. I also hear that when you do this you can tighten your timings. What should I tighten mine to? Right now I have mine at stock EPP 5:5:5:15 (2T) at 800MHz (it was at 1066MHz and 3:4 FSB:DRAM ratio until I bumped it back to make it 1:1). Do I try 4-4-4-12 (2T) at 800MHz or should I put it back to the 1066MHz memory speed? I just want to get my money's worth in performance and stability. Thanks.
 

Mondoman

Splendid
Jun 22, 2006
5,678
0
25,810
5
Neither change will make any noticeable difference in performance. You can try running at DDR2-800 with CL4, but you'll need to do a couple of day's worth of memtest86+ and Orthos (or similar PRIME95-based) stress-testing to be able to trust your system again. Make sure you have set your DIMM/RAM voltage to the spec value recommended by Corsair for your particular DIMMs (should be printed on the DIMMs).


 

mikeynavy1976

Distinguished
Feb 14, 2007
452
0
18,810
16
I ran Memtest (1 pass) on both settings and they passed. My voltage is at 2.13V (spec value is 2.1V although my mobo has Vdroop so I needed to bump it up a bit to ensure at least 2.1V). I've decided for now to leave it at 3:4 FSB:DRAM ratio because of Memtest numbers. At CL4 and 800MHz my memory bandwidth was showing just over 4800MB/s. At CL5 and 1066MHz the bandwidth was showing just over 5200MB/s. I didn't notice any speed increase either way but I figured I'd leave it at a manufacturer setting. If there is a good benefit to using 1:1 ratio and 800MHz at CL4 I'm all ears and wouldn't have a problem switching it back. Please advise.
 

Mondoman

Splendid
Jun 22, 2006
5,678
0
25,810
5
It's just that the memory bus bandwidth boost doesn't matter, since all the data has to go through the FSB, which is the bottleneck. That's why you aren't seeing any performance difference. It certainly won't hurt anything to leave it at DDR2-1066, though.
 

mikeynavy1976

Distinguished
Feb 14, 2007
452
0
18,810
16
Ok...that's what I thought. Would putting it to 800MHz (1:1) ratio and tightening the timing be better though? What about long-term stability/reliability?
 

Mondoman

Splendid
Jun 22, 2006
5,678
0
25,810
5
In theory, yes, but I doubt you'll see any difference. There shouldn't be any stability or reliability issue either way, as the RAM and MB are operating within their designed conditions.
 

mikeynavy1976

Distinguished
Feb 14, 2007
452
0
18,810
16
What benchmark should I use? I just did memory benchmarks with Everest and got the following values:
@CL5 and 1066MHz
Memory Write = 7453 MB/s
Memory Read = 8499 MB/s
Memory Copy = 8413 MB/s
Memory Latency = 71.7ns

@CL4 and 800MHz
Memory Write = 6531 MB/s
Memory Read = 7453 MB/s
Memory Copy = 8158 MB/s
Memory Latency = 84.7ns

This makes it look like I should use the CL5 at 1066MHz. Is there another benchmark I should download and use that shows the system performance difference between 1:1 and 3:4? One thing I did notice is that at 1:1 my motherboard CPU voltage lowers from 1.3 to 1.2? Why does the CPU voltage have any dependency on memory settings?
 

Similar threads


TRENDING THREADS