DDR2 PC-6400 timings help 4-4-4-12 5-5-5-12 & 18.

andrern2000

Distinguished
Jul 28, 2010
507
1
19,015
I plan to buy 8 GB memory.

Is there significant difference between memory timings 4-4-4-12, 5-5-5-12, and 5-5-5-18?
The price for CAS4 is expensive, about $120 or so.
5-5-5-12 is $30 cheaper, and 5-5-5-18 is about the same as 5-5-5-12.

Now I have 2 GB of that 5-5-5-18 one.
If the difference between 4-4-4-12 vs. 5-5-5-18 is noticeable, I might buy the CAS4, considering this is the last upgrade so I wish to expect an optimized buy and performance.

On the other hand, my mobo supports 1066 OC, so should I buy DDR2-1066 instead? But the price is about the same as 800 4-4-4-12.
 
Solution
andren, I did a lot of testing on my first C2 system (eVGA 680i and E6600).

The difference between 4-4-4-12 and 5-5-5-15 at 3.3 GHz (only had an AC7FP at the time) was about 3% in memory i/o. It had no effect on system benchmarks.

The RAM was DDR2-1000 Crucial Ballistixs. I eventually worked the timing from 5-5-5-15-2T down to 3-3-3-7-1T. 6% improvement in memory i/o. Again, no effect on system benchmarks.

Because performance does not scale linearly with FSB freq, higher FSB/lower internal multiplier helps little also.

My recommendations: set FSB:RAM to 1:1 at stock freq, then do what you need to do reach your overclock limits while maintaining stability and staying under Intel's voltage and thermal limits.

And don't rely on Auto...
I OC only to match FSB ratio to 1:1. My CPU is Q6600 I plan to OC it to "400*6 with 1600 Mhz" so that i can fully utilize all PC-6400 benefits.

How's that? So does tighter timings have 5% effect in that condition? If so, I'll take it.

Cheapest option means I'll just buy more of current sticks. But it's brand is V-Gen, a memory directed towards market of value, with 5-5-5-18 timings.
Is that worth it to replace and change into high-class brand like Patriot? Patriot offers 5-5-5-12 at cheap price.
 
Well,..

To get a stable OC (light) to 2,8 Ghz 1600 Fsb,.. so that it achieve 1:1 ratio with DDR2 800.

I think it almost no need to adjust the voltage because it's still light, right? Just leavin' it to Auto BIOS.

But turning off the RAM speed from Auto to manually set 800 so that it won't go past 800 caused by the "Auto" setting..

Anyway, thanks for the info.
 
andren, I did a lot of testing on my first C2 system (eVGA 680i and E6600).

The difference between 4-4-4-12 and 5-5-5-15 at 3.3 GHz (only had an AC7FP at the time) was about 3% in memory i/o. It had no effect on system benchmarks.

The RAM was DDR2-1000 Crucial Ballistixs. I eventually worked the timing from 5-5-5-15-2T down to 3-3-3-7-1T. 6% improvement in memory i/o. Again, no effect on system benchmarks.

Because performance does not scale linearly with FSB freq, higher FSB/lower internal multiplier helps little also.

My recommendations: set FSB:RAM to 1:1 at stock freq, then do what you need to do reach your overclock limits while maintaining stability and staying under Intel's voltage and thermal limits.

And don't rely on Auto settings
 
Solution