2-2-3-3 or 2-2-3-7; int or float. Which is better?

[VIAN]

I have 2 sticks of 512MB PC2700 Corsair Value Select.

I was able to overclock it to 190 and set the timings at 2-2-3-3. is there something wrong with this. No one else sets it at these timings. It runs stable under load for at least four hours. That's enough for me. It won't be running under load for much longer than that. I never turn my computer off and it's been running for about 2weeks with no problems.

If you want to know perforance. In SISandra, memory bandwidth test, it tests similar to kt400 running DDR 400.


Int Buffered aEMMX/aSSE 2863 MB/s
Float Buffered aEMMX/aSSE 2668MB/s

 

[pillage2001]

You're on a Nforce2 board. That's why your memory scores are higher than regular Kt400 boards due to dual channel. Many would prefer raw FSB speed rather than timings as it does not give as much of a performance boost as when you have it at higher FSB.

 

[VIAN]

My Nforce2 doesn't have dual channel. This is single channel operation. ASUS A7N8X-X, check it out.

 

[pillage2001]

Originally posted by: VIAN
My Nforce2 doesn't have dual channel. This is single channel operation. ASUS A7N8X-X, check it out.

Different chipset, different set of scores.

 

[Duvie]

As far as everything I have read, seen tested by respected members, and seen stated at a few websites is the last timing should be equal to the first 3 added together....Anything less then that doesn't result in better performance....

The reason you don't see ppl running less then that is cause they know this....

Should be cas 2,2,3,7....pretty much what I run on my dual channel i875 mobo.....

 

[VIAN]

But why then, do some manufacturers go below that amount.

2-3-3-6
2-3-2-5
2-2-2-5

 

[THUGSROOK]

what?...
you think manufacturers know wtf their doing? think again

 

[Duvie]

ask the guys who tested 2,2,2,5 and found 2,2,2,6 was actually faster!!!

I do see manufacturers doing a hair below that...I just saw a bunch of cas 3,3,3,8 and cas 2.5,3,3,7...so generally you see they are close but nothing like yours....

YOu should test it out using aida, sissoft, and memtest and see if actually bandwidth increases or not...see if theree is a penalty due to possibly having the timings synched that way....

 

[VIAN]

I'll try 2237.

 

[VIAN]

OK, I tried them out. Ran 5 tests in SISandra for each timing then found the average.

2-2-3-3

int- 2850.2
fl- 2630.8

2-2-3-7

int- 2848.6
fl- 2633.2

It comes down to which is more important int or fl. I think fl is more important. What are your thoughts.

 

[Duvie]

I am not sure......Not much of change though, huh??? I would try looking at bandwidth with memtest or even aida32...I never really like sisofts memory test anyways or any of their test as their numbers never seem to tell the whole story...thugs I think used a bunch of them as well as UT2003dome and a few other games.....