Sign in with
Sign up | Sign in
Your question

OC FSB question

Last response: in Overclocking
Share
May 3, 2007 12:47:05 PM

If I scale my settings appropriately and crank up the FSB to get the favorable 1:1 memory ratio what do I have to monitor thermally? Will the the NB get the bulk of the heat or something else? Anything else I need to adjust besides the multiplier, mem/fsb ratio and locking the PCI bus settings? I don't plan on bumping the voltages any.

Or, is the 1:1 ratio not that big of a deal?

I'm targeting 2.92g @ 417fsb(x7), 834 @ 4-4-4-12 (1:1)
(I've run the memory at those settings w/o any problems)

I run 2.92g @ 325fsb(x9), 812 @ 4-4-4-12 all day right now.

It's an
Asus P5B-E
E6600
2x1g Patriot, 800@4-4-4-12
eVGA 8800GTS 320
WD 75g Raptor

Thanks

More about : fsb question

a b } Memory
May 3, 2007 1:37:05 PM

um try the x8 multiplier so it's 3.2GHz with 400FSB and 800 memory. Thats a good config.

The 1:1 allows the memory to run in sync with the FSB and causes less latency from what I hear. it just takes one more step out of the process.
May 3, 2007 2:25:07 PM

please post benchmarks before and after. I'd be curious if there is any difference if all you're changing is the fsb. Show me some mp3 encoding/decoding benches.
Related resources
May 3, 2007 4:09:23 PM

Quote:
um try the x8 multiplier so it's 3.2GHz with 400FSB and 800 memory. Thats a good config.

I want to run no more than 3g just because I have the stock cooler.

Quote:
please post benchmarks before and after.

Running 3g @ 334fsb, 835 @ 4-4-4-12 using 3dmark05 I got 17386
(vid card running 650/950 amazing!)

I'll change it tonight and see how much the 1:1 helps
(if the board will run 417fsb reliably)

Any comments on what to monitor for heat?
a b } Memory
May 3, 2007 6:25:02 PM

whether you use the stock cooler or not does not depend on your speed but how much juice you are giving it and what the temps are. You could clock it to 4GHz if the stock cooler keeps it cool enough.

Lowering your multipler to x8 and you wont need a 417 bus for 2.92ghz at x7. just a thought to play around with.
May 4, 2007 12:15:08 PM

Did some testing last night and unfortunately I found that I can't do an exact comparison because the scalers just don't line up.

My reference was:

2.92g / 417fsb(x7) / 835ram 17026 / 10639 (3dmark05 / cpu)

All I could do was:

2.93g / 325fsb(x9) / 815ram 17110 / 10933

or

3.00g / 334fsb(x9) / 334ram 17391 / 11123

I need a different benchmark(free :) ) to compare ram speed only so these results don't have much meaning. If I was able to match up the speeds and the cpu marks were the same then the 3dmark05 value would have been a reflection of the ram speed, but I couldn't do that.

What is really strange with the results is that the cpu result was about 2.7% slower with the x7 multiplier. there must be some inefficiency at that ratio.

Quote:
whether you use the stock cooler or not does not depend on your speed but how much juice you are giving it and what the temps are. You could clock it to 4GHz if the stock cooler keeps it cool enough.

I'm as hot as I's like to go at 2.93g, maybe I'd push it to 3.0g if necessary but anything over that will surely overheat the proc. All things being equal increasing the frequency will increase the heat generated. The majority of the heat generated is at the level transition, if you increase the frequency there are more transitions per unit time therefore more heat generated.
!