How does reducing the multiplier affect strain on the CPU

[LOUISSSSS]

As the title asks...

and,

why does lowering the multiplier allow us to OC the FSB higher? (not taking ram into account, assume it is running at 1:1)

for example Q6600 Stock = 266 x 9 = 2.4ghz

9 x 400 fsb = 3.6ghz (pretty hard to achieve for most)

8 x 400 fsb = 3.2ghz (not as hard to achieve for most)

 

[SpeedEng66]

if your thinking cpu's only have a fsb wall it's true!

but each cpu/generation has a general max limit of how far that cpu will go (speed wise)
(unless you go with extrem cooling/measures to squeeze out the last ounce of speed)

think about it.. whats the average oc for a conroe?
why can a e2160 oc to 3.2+/-
why can a e6750 only oc to 3.2+/-

thats pretty much the max speed (+/-)<---

 

[heyheybooboo]

How does reducing the multiplier affect strain on the CPU

I'm not sure - but now you kinda have me worried -

As long as your voltages and temps are under control I would say little if any. But if you have to crank +.25v at x8 as opposed to x9 at 400fsb I'd say it creates 'strain' - can't really say if that is a problem though if you are running under temp & volt spec ...

I use CrystalCPUID to manipulate my multi's and voltages. Even though I may up the fsb 25%, depending upon cpu utilization it will kick back the multi and voltage (like SpeedStep, PowerNow and C&Q). It gives you ALOT more options ...

 

[Billb2]

FSB x Multiplier = CPU speed, so:
600 x 5 = 3ghz
300 x 10 = 3ghz
200 x 15 = 3ghz

As far as CPU speed is concerned, it doesn't matter what combination you use.

Where the confusion starts:
CPU power conservation schemes will automatically change multipliers.
Some CPU's are multiplier locked (you can only lower the multi from stock, not raise it).
Motherboards have a limit of how high a FSB they can run.
Memory speed is also controlled by FSB.
Some benchmarking programs benefit from a higher FSB at the same CPU speed (ie. 600 x 5 will bench better than 300 x 10)

 

[LOUISSSSS]

so if my CPU is having a hard time reaching 400mhz fsb, (9x multipler = 3.6ghz) then it will have that same problem reaching 400mhz fsb ith an 8x multiplier?

 

[Denithor]

Generally it's not so much your cpu having a hard time with 400MHz it's your motherboard. Running higher fsb/lower multiplier versus lower fsb/higher multiplier makes no difference to the cpu (500x6 versus 333x9) because the operating speed is the same. However, running the higher fsb is harder on the motherboard and generates more heat that has to be dissipated.

 

[LOUISSSSS]

^^ and this heat will be generated from the NB right

 

[geokilla]

I wonder about this as well. How does overclockers get a higher FSB/HT speed when lowering their multiplier?

If I leave my multiplier at 10, my FSB tops out at around 240 before getting BSOD. If I drop it to 9, I could push it up to around 280+.

 

[Markfw]

In some cases, its the motherboard that has a fsb wall, other times its the CPU. My ASUS P5K has a crappy FSH and power delivery, my DS4 has a FSB wall of 509, but my X3350 won;t go past 450.

 

[bryanW1995]

Originally posted by: SpeedEng66
if your thinking cpu's only have a fsb wall it's true!

but each cpu/generation has a general max limit of how far that cpu will go (speed wise)
(unless you go with extrem cooling/measures to squeeze out the last ounce of speed)

think about it.. whats the average oc for a conroe?
why can a e2160 oc to 3.2+/-
why can a e6750 only oc to 3.2+/-

thats pretty much the max speed (+/-)<---

what??? an e6750 will only oc to 3.2 +/-? most e6750's that I've seen/heard about get to 3.2-3.4 on stock voltages, then go up from there. I run mine at only 3.4 b/c I'm running seti 24/7 and the heat bothers me, but I can easily run mine higher and I have an average chip imho. I haven't heard of a single e6750 that is limited to 3.2, while I've only rarely seen an e2160 get that high.

 

[betasub]

Originally posted by: bryanW1995
what??? an e6750 will only oc to 3.2 +/-? most e6750's that I've seen/heard about get to 3.2-3.4 on stock voltages, then go up from there.

Believe it. Not every CPU was born to overclock, and with so many Conroes giving stellar overclocks, we kinda forget about the few that don't (even though they still make reasonable OCs by any other standards).

 

[Mango1970]

I pay extra for the higher multiplier -- personally I have better "luck" getting a really good OC using a LOWER FSB and higher X. However this makes me play much more with the RAM as usually 1:1 is crap and you find yourself underclocking your RAM with that setting or you need to start playing with different straps etc. Hence why I still love my 680i and running async to set my ram at whatever I want. However I also love my Asus P5Ke-WiFi and Ballistix RAM since lowering your timings gets you higher "clocks" on the RAM... generally I would rather have my ram run lower timings but it's a trade off.

 

[SpeedEng66]

Originally posted by: bryanW1995

Originally posted by: SpeedEng66
if your thinking cpu's only have a fsb wall it's true!

but each cpu/generation has a general max limit of how far that cpu will go (speed wise)
(unless you go with extrem cooling/measures to squeeze out the last ounce of speed)

think about it.. whats the average oc for a conroe?
why can a e2160 oc to 3.2+/-
why can a e6750 only oc to 3.2+/-

thats pretty much the max speed (+/-)<---

what??? an e6750 will only oc to 3.2 +/-? most e6750's that I've seen/heard about get to 3.2-3.4 on stock voltages, then go up from there. I run mine at only 3.4 b/c I'm running seti 24/7 and the heat bothers me, but I can easily run mine higher and I have an average chip imho. I haven't heard of a single e6750 that is limited to 3.2, while I've only rarely seen an e2160 get that high.

average

thats pretty much the max speed (+/-)<---

im not sayin it cant hit 3.5+ but average oc

please read

 

[aigomorla]

reducing multi puts strain on the rest of your components.

The NB now has to work harder, your ram is also clocked harder, and more voltage overall is required for the higher FSB.

We have been honestly spoiled by intel. Because of the decient multi's on the C2D, we really didnt need uber fast ram. Thats why no one really bought DDR3.

Now were getting crap multi's 8.5x highest on a 500 dollar quad. And 9.5x on a 400 dollar Wolfdale

we're back in the AMD days where 75% of the failed OC settings was due to memory failure.

Louis in terms of speed, there isnt much of a difference compared to 9x400 vs 8x450.

Just try to keep your fsb in the 400 strap by getting as close to as possible or over 400fsb.

 

[T2k]

Originally posted by: LOUISSSSS
9 x 400 fsb = 3.6ghz (pretty hard to achieve for most)

People keep repeating this - is there any truth in it? Reason I'm asking is that all Q6600 I came across - owned, tried, heard of etc - were able to do 3.4-3.5GHz with stock cooler, with no or very little extra Vcore. Even some ghetto mobo like mATX Asus with zero OC features runs one here @3.4GHz in a silly little hp dx5150 case, with stock cooling. My former home Q6600 did 3.6GHz pretty easy, anytime, with a little extra Vcore - it was 3.8GHz and beyond when it became pigheaded and required some serious voltage.
I wonder if it's true and what batches we're talking about - the ones I am talking about are all last Nov-Dec ones.

 

[T2k]

Originally posted by: aigomorla
We have been honestly spoiled by intel. Because of the decient multi's on the C2D, we really didnt need uber fast ram. Thats why no one really bought DDR3.

Now were getting crap multi's 8.5x highest on a 500 dollar quad. And 9.5x on a 400 dollar Wolfdale

we're back in the AMD days where 75% of the failed OC settings was due to memory failure.

This argument doesn't make any sense to me - any half-decent mobo could utilize lower memory speed by using dividers since NF2 IIRC. Why do you think it's the RAM?
No, IMO DDR3 was/is expensive as hell and sports a higher latency, that's why it haven't really made inroads.

 

[aigomorla]

Originally posted by: T2k

This argument doesn't make any sense to me - any half-decent mobo could utilize lower memory speed by using dividers since NF2 IIRC. Why do you think it's the RAM?
No, IMO DDR3 was/is expensive as hell and sports a higher latency, that's why it haven't really made inroads.

Okey you try clocking a wolfdale at 500FSB on DDR2 667/800 Ram.

No multi will save you from this.


Unless your running 1066's or Micron D9's good luck!

 

[Billb2]

Originally posted by: LOUISSSSS
so if my CPU is having a hard time reaching 400mhz fsb, (9x multipler = 3.6ghz) then it will have that same problem reaching 400mhz fsb ith an 8x multiplier?

FSB, multipliers and dividers........ Oh, you're confused!

FSB is limited by your motherboard, it will only go so high.

Since CPU speed is FSB x Multiplier, there are two limits 1.) How high a FSB the Mobo will run and 2.) How fast the CPU will run. So long as you're below the Max FSB for the motherboard and the Max speed for the CPU, any combination of FSB and Multiplier is fine. So, is your FSB to high for the motherboard or is the CPU speed to high?

Since Memory speed is FSB x divider, there are two limits 1.) How high a FSB the Mobo will run and 2.) How fast the memory will run. So long as you're below the Max FSB for the motherboard and the Max speed for the memory, any combination of FSB and divider is fine.

Overclocking without first determining these three maximums (Motherboard FSB, CPU speed, and memory speed) independent of each other is sheer folly. You get to where you are now. You don't know if the FSB is to high or is the Multiplier is to high and you're probably going to have the same problem when you memory overclocks fail.

 

[aigomorla]

bill you forgot that some boards also have FSB walls. So theres sometimes a range where it wont boot up for dog dookie.

I remember i had a wall from 375-410. Anything in those range, the board wouldnt boot up. If i was below, or above, the system ran perfectly.

 

[GuitarDaddy]

The simple answer to the OP's question

"How does reducing the multi affect strain on the CPU?"

It reduces the CPU speed!!!


In the example you gave 9x400 vs 8x400, the stress on the mobo and the ram are the same in both schenarios, the difference is 3.6ghz vs. 3.2ghz. Obviously it's more difficult to get the CPU running stable at 3.6ghz, it is that simple.

A better question might be, "How come my CPU will do 8x455=3640 on the 8x multi, but can only do 9x400=3600 on the 9x multi" Not saying that's the case with your chip, but I have seen examples of this schenario

 

[T2k]

Originally posted by: aigomorla

Originally posted by: T2k

This argument doesn't make any sense to me - any half-decent mobo could utilize lower memory speed by using dividers since NF2 IIRC. Why do you think it's the RAM?
No, IMO DDR3 was/is expensive as hell and sports a higher latency, that's why it haven't really made inroads.

Okey you try clocking a wolfdale at 500FSB on DDR2 667/800 Ram.

No multi will save you from this.

But that's not the memory - I can keep memory running as low as I want. It's either the chipset that cannot supply 500MHz or the CPU that cannot take it.

Unless your running 1066's or Micron D9's good luck!

WEll, I actually DO run D9xx PC6400 modules (see config link in sigline).

 

[aigomorla]

Originally posted by: T2k

Originally posted by: aigomorla

Originally posted by: T2k

This argument doesn't make any sense to me - any half-decent mobo could utilize lower memory speed by using dividers since NF2 IIRC. Why do you think it's the RAM?
No, IMO DDR3 was/is expensive as hell and sports a higher latency, that's why it haven't really made inroads.

Okey you try clocking a wolfdale at 500FSB on DDR2 667/800 Ram.

No multi will save you from this.

But that's not the memory - I can keep memory running as low as I want. It's either the chipset that cannot supply 500MHz or the CPU that cannot take it.

Unless your running 1066's or Micron D9's good luck!

WEll, I actually DO run D9xx PC6400 modules (see config link in sigline).

Heh.. T2K trust me bro. When i was playing with a wolfdale i was glad i busted out my 1066 tracers.

on the Wolfdales and yorkie's i promise your gona run into memory speed problems as you get closer to 4ghz and expecially as you pass it.

 

[T2k]

Originally posted by: aigomorla

Originally posted by: T2k

Originally posted by: aigomorla

Originally posted by: T2k

This argument doesn't make any sense to me - any half-decent mobo could utilize lower memory speed by using dividers since NF2 IIRC. Why do you think it's the RAM?
No, IMO DDR3 was/is expensive as hell and sports a higher latency, that's why it haven't really made inroads.

Okey you try clocking a wolfdale at 500FSB on DDR2 667/800 Ram.

No multi will save you from this.

But that's not the memory - I can keep memory running as low as I want. It's either the chipset that cannot supply 500MHz or the CPU that cannot take it.

Unless your running 1066's or Micron D9's good luck!

WEll, I actually DO run D9xx PC6400 modules (see config link in sigline).

Heh.. T2K trust me bro. When i was playing with a wolfdale i was glad i busted out my 1066 tracers.

on the Wolfdales and yorkie's i promise your gona run into memory speed problems as you get closer to 4ghz and expecially as you pass it.

We'll see. However I bet it will be either the CPU or the mobo that won't be able to hold the line, not the memory - I don't even introduce memory OC until I found the highest stable FSB/multiplier combo, just keep uit on base speed though it's true that quite often BIOSes are crappy and crap out when asynch dividers are set...

 

[Billb2]

Originally posted by: aigomorla
bill you forgot that some boards also have FSB walls. So theres sometimes a range where it wont boot up for dog dookie.

I remember i had a wall from 375-410. Anything in those range, the board wouldnt boot up. If i was below, or above, the system ran perfectly.

That's limited to a few chipsets.
Northbridge buss frequencies don't ramp up with FSB but step up in jumps. Some of which render unstable conditions. So, yeah, there can be "holes" where overclocks fail.