[Villainstone]

Well guy I am a bit confused about whats going on. Like anyone here I am looking to get the most out of the rig I own. I decided to play with the FSB, multiplier, and RAM divider. I am looking to get the fastest over all real world performance so this it what I did.

For the longest time I have been running my rig at 3.6 on the CPU (400x9), and 1000 (500mhzx2). Obviously was in a 4:5 divider and started thinking about a 1:1 so here is where I started scratching my head. I went from what I listed above to 450x8 (still 3.6) and RAM to 900 (450x2). Basically RAM is slower, FSB is higher plus I am in a 1:1 divider... make sense right. So at 3.6 and 900mhz in a 1:1 I ran ~ the same Super PI as the original OC (~14.5sec). Next I decided to try 500 FSB in a 1:1 divider and see where that puts me at. So here is where I am at 3.5ghz CPU, DDR2-1000mhz 4-4-4-5, and 500mhz FSB and still running the same Super PI times. I am saying whats the deal... three different settings and all withing milla-seconds of the same times.

So if I want the fastest real world performance and all the settings run the same times, then what should I set it at and why. Shoud I stay at the 3.5, 1000, and 500 FSB 1:1 or the original stetting which call for much less voltage on the NB? I am a bit in the dark here so please let me know what you think.

 

[Modki]

Running the RAM faster than the CPU does nothing for performance.

Best bet would be to run it at 400Mhz, 9 Multi, 1:1 RAM Ratio.

 

[Villainstone]

Right I can see that but why isn't 500Mhz x7 and 1000Mhz in a 1:1 faster than that in Super PI times I just don't get it. I want it all to run fast with high Mhz lol.

 

[Modki]

I'm not sure why you want one 3.5Ghz to be faster than another 3.5Ghz...

 

[Nevaziah]

EEK !!!

Modki's back. Repent, repent for I say the world's ending is at hand!!

Ahem*
Technically the SUPER PI only AFFECTS the cpu computing speed. try running other benches like 3dmark06, Sandra or 3dmark05 to see if 1:1 is better than 4:5.

 

[Villainstone]

Quote:

Originally Posted by Modki I'm not sure why you want one 3.5Ghz to be faster than another 3.5Ghz...

No no no... I am looking to see which setting is faster. Ohh just nevermind.

It's so hard to find good honest help around here these days. Maybe I give up too soon anyway thanks again.

 

[Modki]

No honestly, you're just using SuperPi. 100Mhz isn't going to make that much of a difference. Running the RAM faster than the CPU means you're just stressing the RAM more than you need to because it won't increase performance.

Honest.

Go try 3Dmark06 now. It'll give you like a 100 point difference between them, honestly.

 

[pauldovi]

Quote:

Originally Posted by Modki Running the RAM faster than the CPU does nothing for performance. Best bet would be to run it at 400Mhz, 9 Multi, 1:1 RAM Ratio.

You are very wrong. Running ram faster than a "1:1 ratio" is beneficial.

Running your memory "out of sync" or in a "divide" does not affect performance at all. Having a 4:5 ratio is perfectly fine. Try to get both your FSB and your memory speed as high as they can go. Ignore the whole "divider" thing. It only serves to confuse people.

 

[Villainstone]

Quote:

Originally Posted by Modki No honestly, you're just using SuperPi. 100Mhz isn't going to make that much of a difference. Running the RAM faster than the CPU means you're just stressing the RAM more than you need to because it won't increase performance. Honest. Go try 3Dmark06 now. It'll give you like a 100 point difference between them, honestly.

TY Modki thats better.

Quote:

Originally Posted by pauldovi You are very wrong. Running ram faster than a "1:1 ratio" is beneficial. Running your memory "out of sync" or in a "divide" does not affect performance at all. Having a 4:5 ratio is perfectly fine. Try to get both your FSB and your memory speed as high as they can go. Ignore the whole "divider" thing. It only serves to confuse people.

That is so true in my situation I just feel like the divider and a 1 digit difference in timings don't really mean much. (4-4-4-5, to 4-4-4-4). On the other hand like you guys said I am only running S.PI so I will go and get a sull version of sandra and see what I get. Thanks and REP to you both.

 

[Taeric]

Quote:

Originally Posted by pauldovi You are very wrong. Running ram faster than a "1:1 ratio" is beneficial. Running your memory "out of sync" or in a "divide" does not affect performance at all. Having a 4:5 ratio is perfectly fine. Try to get both your FSB and your memory speed as high as they can go. Ignore the whole "divider" thing. It only serves to confuse people.

I think that is currently a common misconception based on older tech, which did not benefit much if at all from running a divider that favored the RAM. There is a significant amount of documentation that shows real world benefits from dividers on current tech, so hopefully that misconception will die out over time.

 

[Modki]

Ah I didn't know that see. So running the RAM higher than 1:1 is better now?

 

[Taeric]

It has a definite potential to increase performance. As always, though, the best way to tweak things out is good ol' trial and error, especially if you need to loosen timings to achieve higher frequencies. Some applications respond to one better than the other, but either way, it should definitely be a part of your testing process if your RAM has room to OC beyond what the CPU can.

 

[MADMAX22]

I guess it all depends on what your doing, I know that my best performance comes from getting the fsb as high as stabily possible. I get better scores running 500x8 1:1 then I do with 445x9 1:1. and thats in pretty much every benchmark I run.

Also if and when I can I run a faster ram speed as that gets better bandwidth obviously and better performance in most benchmarks. When I run 445fsb I run a 4:5 divider which makes better performance atleast in my case.