Ram Timing vs higher clock speed

[xhail]

I have a Intel 2.4C running on an ASUS P4P800 deluxe with 512megs of Geil PC3200.

I can get 240FSB using Ram times of 2.5/8-4-4 in the bios and I can run it at 200FSB (stock) at 2/6-3-3. Is it better to just leave it at 2.4G and running those timings? or is it better to get 240FSB with lax timings? Is there some sort of rule of thumb as to mhz increase vs lax timings for RAM?

p.s. I mostly play vid games, capture tv and encode music/dvds

Thanks

edit: I am running DDR voltage at 2.85 (max the P4P800 will go)

 

[heyheyhey]

You won't even be able to even begin to see any difference. Low RAM timings are more for bragging rights, unless its like 20-15-15-10

 

[jmcmike]

Back when I put my current system together I did some tests on the subject and found that timings matter whole lot, but not at the expense of CPU speed.

http://hardforum.com/showthread.php?s=&threadid=479609

At the same or close FSB speeds (same or close CPU OC) I have found that timings make a big difference. Especially if you can tune your divider settings more precisely than I can with this board.

In a nutshell, I got the same performance with a 1:1 divider @ 340 DDR fast timings than I did with a 3:4 divider @ 444 DDR slow timings. Upon further testing I found that timings basically rule, so long as you're not losing CPU overclock. I can only imagine that with the closer divider settings of the new boards you could really get the most out of your RAM with fast timings.

In my case, since the performance difference between the two was neglegible, I decided to keep the higher speed with slower timings simply for the coolness factor of a bigger number for my DDR speed.

 

[xhail]

thanks for the info

Now I got one more question. How much of an overclock justifies using a 5:4 divider on a P4 2.4C?

I read somewhere that the performance goes way down if you don't keep the FSB 1:1. But I'm sure there is some sort of magical figure that would determine when one should actually use a 5:4 divider.

Would an overclock of 2.4 @ 3.2 justify in terms of performance a divider of 5:4? Or should that never be used except to show off?

Thanks

 

[Swat]

Originally posted by heyheyhey
You won't even be able to even begin to see any difference. Low RAM timings are more for bragging rights, unless its like 20-15-15-10

Well , I've ran a few bench marks with timmings being very different. Each time the same resulted in an average of about 10+ fps. The lower timmings always beeing faster.

 

[NEVERLIFT]

In testing it shows that a Cas2 vs Cas3 is like a 30-50mhz fsb increase and it only gets better with lower timings.

2-2-2-11/2-2-2-6 are the timings you want

I would always go for this over a higher fsb and slower timings.

 

[jmcmike]

Originally posted by xhail
Now I got one more question. How much of an overclock justifies using a 5:4 divider on a P4 2.4C?

To my knowledge there is no formula that can answer your question. Just benchmark each scenario and pick the one that's the fastest.

 

[Nitro16]

Generally, going from 1:1 ratio to 5:4 ratio at the same timings causes around a 2-4% drop in performance, basically nothing.

Running slack 2.5-4-4-8 timings instead of tight 2-2-2-5 timings results in about 3-6% drop in performance. Usually 2-2-2-5 at 5:4 ratio actually beats 2.5-4-4-8 in performance slightly.

On Intel i865/i875 chipsets CAS latency has a very minor effect on performance. The timing which affects it the most is the RAS-to-CAS delay, going from 2 to 4 and you basically drop anywhere around 3-7% in performance on the spot. This is why BH5 is highly sought after instead of CH5 (aside from compatibility issues). Most CH5 can only do 2-2-3-5, where as BH5 can do RAS-To-CAS 2 (2-2-2-5) equalling slightly higher performance.

 

[xhail]

thank you very much for the information!!