Why would you lower your multiplier on OC?

[mhouck]

Ok. I've done the mandatory google searches for my Q6700 and nearly every overclocking article shows lowering the multiplier. I'm assuming that it has to do with the memory dividers but I don't understand why you would lower your multiplier below 9x and 400 fsb. Look at this article for instance

Hexus- Core 2 Quad GO-stepping overclocking-Q6700 retail

"With core voltage pushed up to 1.5V in the BIOS, which yielded around 1.45V under full load, we were able to get the Q6700 stable at 3.6GHz. The ASUS P5K Deluxe motherboard also managed to produce the goods, hitting the 1800MHz (450MHz QDR) FSB mark with ease.

Pushing even further, we were able to run benchmarks at just over 3.7GHz, complete with a whopping 1852MHz (463MHz QDR) front-side bus."

These fsb clocks seem a bit ridiculous(sp?) considering it has two steps higher on the multipler to go still. What's the point of getting the bus so high?
Can it possibly be better for the northbridge? Is it for stablity reasons? If not would should I use the 10x multiplier and just chaning the memory dividers to something close to my memories rated speed?

My signature contains my system and I have a DFI LANPARTY DK X48-T2RS due to arrive on Wednesday to replace my current board.

 

[Andrew1990]

I believe they are trying to find the FSB wall on the motherboard when they lower the multi.

 

[mhouck]

Originally posted by: Andrew1990
I believe they are trying to find the FSB wall on the motherboard when they lower the multi.

that seems reasonable but the review is for the processor not the board.
So there isn't an advantage to lowering the multiplier as far as stablity?

 

[Idontcare]

A long time ago conventional wisdom was established that you should attempt to hit both your CPU's maximum overclock and your NB's maximum FSB clock and your ram's maximum clocks (at any crappy timings).

And so it came to pass that if your chip max'ed out at 3.6GHz but it has a 10x multi then you were wasting FSB by leaving the multi at 10x and the FSB at a "paltry" 360MHz. Instead it became quite the rage to decrease your multiplier, say to 9x, and increase your FSB ever more until you reached the your CPU's limit and FSB limit (say at 400MHz on 9x multi).

Then you are supposed to overvolt the crap out of your ram so you can use ram multiplier and have your DDR2 run at 1100Mhz.

So it is not done for stability, and it is not even really done to improve performance. It will improve benchmark results. Sisoft loves super clocked DDR2, but good luck finding an application that you use to actually show an improvement in performance.

Same with FSB...higher FSB does increase bandwidth between the cores on a quad, but good luck finding anything but benchmark programs actually registering this.

Increased FSB on single-socket desktop systems is mostly an Intel marketing ploy. Increased DDR2/DDR3 speeds is mostly a ploy for the marketing folks of the dram guys.

There were some threads here a while ago (search for graysky) where folks were posting their FSB and DDR2 clocks versus performance...conclusion was it was a waste of time and just burned up your board faster without giving anything in return.

But it is not as sexy to say you merely overclocked your CPU but kept your FSB and DDR2 speeds to a minimum.

 

[Tweakin]

And the e-penis award goes to......

 

[Tempered81]

some people like 8 x 500 more than 9 x 445 for 4000mhz. the 500mhz FSB speed and faster clocked ram usually result in better system performance, even though the cpu is running at the same speed. 445 to 500 is not as noticable as say 266 to 500 would be.

You also need fast ram to run those clocks and keep decent timings.

Originally posted by: Tweakin
And the e-penis award goes to......

awe?

 

[mhouck]

Originally posted by: Tweakin
And the e-penis award goes to......

lol. Ok, so it looks like I'll clock back to 400x9 and do 1:1 for the mem considering my processor lets me get there as my first goal and then work with the 10x multiplier and memory divider. I think that will be best for the long haul.

War Eagle? Auburn gets robbed in the BCS every year. Congrats on spanking Clemson in the bowl.

 

[magreen]

Idc, just a caveat to your point (I know you like caveats):
On AM2 systems the memory clocks do make a big difference to ystem performance, and not just in benchmarks.

But I understand that your point was made for core 2 systems, which it holds true for.

 

[Idontcare]

Originally posted by: magreen
Idc, just a caveat to your point (I know you like caveats):
On AM2 systems the memory clocks do make a big difference to ystem performance, and not just in benchmarks.

But I understand that your point was made for core 2 systems, which it holds true for.

Oh you get a big ass smiley for that one Caveats FTW!

And yes, naturally, I defer to your stated caveat. It is appropriate and well posited.

I wish I had recent AMD experience to talk from, but alas the K7 (Palomino) was the my last opportunity. Ironically enough when the X2's came out I just couldn't afford them at the time, so I bought a Northwood system with hyperthreading (the poor man's piss-poor dual-core) for my upgrade from the palomino.