2-2-2-11 better than 2-2-2-5?

[Silent Buddha]

Somebody told me 2-2-2-11 is better than 2-2-2-5 for nForce2 chipsets. Is this true? And why?

 

[^wargod]

Yes, for nf2 chipsets it is better. No idea why, but i know it is :]

 

[SeanOMatic]

It has to do with the memory not being able to complete instructions before the cycles is complete.

 

[mrspec3]

All testing and images are by Dogman (crucible.ca)

 

[vincent_1985]

nice charts

 

[Top Hat Theater]

My tests on this topic can be found at: http://www.mods4me.com/forums/showthread.php?s=&threadid=4169

~THT

 

[Silent Buddha]

It has to do with the memory not being able to complete instructions before the cycles is complete.

So this is basically the only reason?

 

[ttfrog]

You should test your ram. It can vary from motherboard to motherboard though 2-2-2-11 is the "established" best.

 

[mfjonny]

Here is a similar thread...

http://www.ocforums.com/vb/vb/showthread.php?s=&threadid=248396&highlight=tras

Though Hoot's findings about multis weren't confirmed outside of his personal testing.

 

[KHysiek]

considering these charts 2226 or 7 looks best, cause it's better simulation of real world usage. You don't use you computer just for Sandra nadwidth benches.

 

[newg]

looks like it was a lucky guess in my going 2-2-2-7..........................

 

[perinator]

Im guessing that these timings are for the nf2 chipsets.

Does anyone know the timings for via kt600 chips?

 

[xgman]

Same for NF3 chipset?

 

[DayUSeX]

hm thnx for those charts and heads up guess its time to tweak my timings again.

 

[Cow2kie]

10x223 2,2,2,13 3360/3152
10x223 2,2,2,12 3364/3153
10X223 2,2,2,11 3371/3162
10X223 2,2,2,10 3367/3159
10X223 2,2,2,9 3357/3152
10x223 2,2,2,8 3353/3147
10x223 2,2,2,7 3352/3150
10x223 2,2,2,6 3351/3147
10x223 2,2,2,5 3340/3138

10X223 2,2,2,11 3371/3162
10x223 2,2,3,11 3356/3145
10x223 2,2,4,11 3354/3128


10X223 2,2,2,11 3371/3162
10x223 2,3,2,11 3355/3136
10x223 2,4,2,11 3340/3110

10x227 2,2,2,11 3436/3224

 

[KHysiek]

Well, looking at above numbers differences are minor, you can get much more from higher FSB with lower timings than from lower FSB with tighter timings on nForce. But It all depends what you expect from your system and what kind of benchmarking Sandra use.
If you want to set records this is the way to go, if you want fastest system I'm not sure if Sandra scores are adequate to real world usage here. In RL apps you more benefit from lower timings (except mass data bandwidth hungry ones like video/graphics editing). I think you should select some typical app which can be used for precise measuring (calculation program, game, 3dmark, 3D renderer) and fire up some runs with different mem settings.

As for strictly TRAS there seems to be just 2-3 options - lowest possible, 6/7 and 10/11.