Lower memory frequency, tighter timings?

[prolfe]

Hi everyone,
I am wondering if it is possible to use memory rated for DDR2667 at 5-5-5-12 and run it at DDR2533 4-4-4-10 (or whatever lower timings I could achieve). It seems logical to me that reducing the frequency could allow you to tighten the timings, but I do not possess enough knowledge of RAM engineering to know for sure. I am sure someone here does, though! I look forward to your responses as I try to pick memory for my new C2D.[/img]

 

[Spiv]

Tightening the timings can introduce instability in the ram. But, since you're downclocking it at the same time, this *should* clear up any instability. The reason I say 'should' is because it's possible that the ram chips on the stick simply can't run at those timings.

With that said, get the ram that you can afford and tweak it (lower timings first, then lower speed if it's needed). Despite what the articles say, it is very difficult to notice differences from memory speed in the real world.

 

[prolfe]

Maybe it would help if you knew where I was coming from here. I've purchased a C2D 6400 (most I could afford). I also got a ECS Nforce570SLI and I need some memory. I have read that timings are more important than frequency with C2D, and that synchronous operation is best if possible. I found some low latency OCZ but they're 533, so if I OC the FSB I don't have much headroom. Since I know I'm not gonna OC the FSB to anywhere close to 667, I was thinking of getting 667 memory with average timings, tightening the timings and running the FSB somewhere around 575. Does anybody think that might work?

 

[SirCrono]

I also got a ECS Nforce570SLI

do you mean the nforce 570 slit-a??, i think they dont support the e6400 (not sure though)

anyway, if you want to get the most of your ram i think you should make it synchronous first and then thigten the timings, that way you'll be sure it's stable (dont forget to test it)

anyway as they'v already said the memory speeds/timings does not make that much of a difference, at least from the human perspective.

 

[rwpritchett]

If benchmarks mean anything to you, I have a little data that comes from doing exactly what you are suggesting.

I have DDR2-667 RAM spec'd for 4-4-4-12 timings. I ran 3dmark05 four times and this is what I got with no overclocking of the CPU:

@DDR2-800 4-4-4-12: 5524 (OC'd RAM, asynchronous)
@DDR2-667 4-4-4-12: 5524 (RAM stock settings, asynchronous)
@DDR2-533 4-4-4-12: 5525 (UC'd RAM, synchronous)
@DDR2-533 3-3-3-8: 5600 (UC'd RAM, synchronous plus tighter timings)

So it appears that tightened timings do have a greater impact than frequency. Not much, but some. This is just a benchmark, however, and the real world difference would probably not be noticable.

 

[prolfe]

Wow!

You guys are incredible. Thank you so much for your quick and informative replies. In response to Crono: Yes it is slit-a, and yes it does support C2D AFAIK; Newegg, at least, says it does. Furthermore, it is revision 1.0B, it may be that the "B" part allows it to support C2D, considering that there IS a model called revision 1. In response to pritchett, your benchies correspond almost exactly to some articles I've read online, so I'm going to go ahead and get the 667 with CL4, instead of 533 CL3. Once I get the rig built I'll try to post some benchies.

 

[prolfe]

Hey everyone what would you think about me switching to the Asus P5NSLI for this rig, instead of the ECS Nforce570SLIT-A?