is the multiplier set by the CPU or the motherboard?

[CTho9305]

Does the CPU have a PLL on it, or does the motherboard's PLL set the CPU frequency? If the PLL is on the CPU, then why does the mobo BIOS have to support the higher multipliers? If it is on the motherboard, why hasn't intel's mutliplier lock been worked around?

 

[Dulanic]

<< Does the CPU have a PLL on it, or does the motherboard's PLL set the CPU frequency? If the PLL is on the CPU, then why does the mobo BIOS have to support the higher multipliers? If it is on the motherboard, why hasn't intel's mutliplier lock been worked around? >>



The PLL has nothing to do with the multiplier, it has to do with the FSB. The MB sets the FSB with the PLL and then the CPU tells the chipset what to set the multiplier to and then the chipset goes back and tells the CPU to run X multiplier. But the CPU can prevent that change also, the CPU can say nope Im running 5x multiplier even if you tell me to run 10x (Intel). The CPU normally just reports the multiplier to the MB and then it sets the multiplier to that (AMD).

 

[CTho9305]

so the motherboard has no real control over the multiplier - something in the core / packaging sets the multiplier.

 

[SCSIRAID]

The CPU contains the PLL that multiplies the FSB clock supplied by the motherboard up to the CPU clock frequency. The multiplier is provided by switches on the motherboard.

Motherboards can also have a PLL that creates the FSB clock and AGP clock.

 

[Dulanic]

<< so the motherboard has no real control over the multiplier - something in the core / packaging sets the multiplier. >>



Right, the chipset can recommend a multiplier, but the CPU itself sets the multiplier and it can say no just as easy as it can say yes

 

[CTho9305]

oh ok . so to overclocked a locked cpu w/o overclocking PCI out of spec, you would need to separate the cpu by another layer from everything.... unreasonable

 

[OS]

<< so the motherboard has no real control over the multiplier - something in the core / packaging sets the multiplier. >>



That's not so, the mobo can and does control the multiplier, but a locked CPU will simply override whatever the motherboard has set.

 

[CTho9305]

.... so we still don't have a consensus

OuterSquare - if the CPU has no PLL, how can it know what the actual multiplier is? it sees a clock pulse... how does it find the multiplier?

 

[johncar]

CTho9305, you wrote.....
Does the CPU have a PLL on it, or does the motherboard's PLL set the CPU frequency? If the PLL is on the CPU, then why does the mobo BIOS have to support the higher multipliers? If it is on the motherboard, why hasn't intel's mutliplier lock been worked around? End quote...

Short answer to your question...Host Clock Multiplier is set by BOTH the CPU AND the mobo...Explanation....

Host Clock is generated by mobo PLL, but that's not the reason we wrote BOTH. Wrote BOTH because the PLL inside the CPU, (which multiplies the Host Clock...really the System Clock but generally also incorrectly &quot;called&quot; the Front Side Bus...clock), gets its multi-bit binary HI/LO multiplier signals code from simple electrical direct current circuits...and on AMD chips these circuits are pre-set by open/closed bridges on CPU exterior...BUT...some mobos connect additional circuits to these original circuits with the &quot;capability to change&quot; the HI/LO polarity of the signals made by the CPU's bridges.

Bios is just one way to make over-riding/overclocking settings in these additional circuits on the mobo...manual jumpers and switches are another way.

So the CPU cannot &quot;say NO&quot; to the altered signals...how could it possibly do so since the CPU PLL responds to the signals' polarity...it doesn't know what where or how the signals were created.

http://members.nbci.com/candjac/index.htm Duron/Tbird OC article illustrates how a person can attach circuits to the default circuits for the purpose of &quot;changing default settings&quot;...just like some mobos do as illustrated on page 3 of the article.
John C.

PS Believe Intel does not accept external signals to its CPU's PLL...they are set/locked &quot;internally&quot;.

 

[CTho9305]

ok, so here is what I understand:
the CPU has a PLL on it. on AMD chips, it is open to input from some pins / easily accessible bridges. on Intel chips, it is set internally and nothing outside can affect it.

does anything on the motherboard actually operate at the cpu internal frequency / do any traces carry clocks at that frequency? or is it really truly internal?

 

[pm]

The motherboard has a clock generator (usually a piezoelectric crystal) which generates a square wave. A PLL (phase-locked loop) on a chip on the motherboard locks onto this signal and generates several other clocks from it that are all locked in phase and these clocks are used to syncronize many of the chips on the motherboard (the CPU, the RAM, the PCI bus, etc). The CPU receives this signal and uses an on-CPU PLL to generate on-chip clocks which are also locked in phase (such as the on-chip FSB and Core clocks).

Intel CPUs (Pentium III, Pentium 4, etc) do have multipliers that can be set by the motherboard (engineering sample CPU's are 'unlocked'), but consumer products are locked to prevent forgeries.

The motherboards may only need to support the multipliers so that they register the frequency correctly. It may be possible to run (for example) an 11x multiplier CPU on a motherboard that doesn't support 11x multipliers but the BIOS will report something like

123MHz CPU detected

on boot up... or some other nonsense. The CPU might still work correctly but it would simply be disconcerting to the uneducated consumer. Having said this, higher multipliers generally also include increased power requirements and an older motherboards may not have been tested to be rated for that CPU's higher power requirements.

 

[CTho9305]

Thanks

 

[Evadman]

So we have &quot;in-step&quot; pulses being sent out by the CPU and Clock gen, with the clock gen being the &quot;master&quot; for the system. All the Pulses are &quot;stepped-up+ from the single clock gen. That is why when you change the clock gen ( Bus Speed adjustment ) everything else changes with it. The CPU sees this pulse, and ajdustes it's internal pulse to be say 10x times it. ( for a 100MHz 1.0 gig proc ) The CPU gets the 10x from both the Motherboard and the CPU. The CPU has the final say until it is &quot;unlocked&quot;.

Now for a kinda related question:

Ok, so we knoe where the multiplyer bridges are on a AMD CPU, but where are they on a Intel CPU? Are they set externaly, or internaly? AMD's would be external, because even though they are on the chip, the enterprising hardware guru can change them. The Slots for AMD are harder, because you have to take apart the case, but they can still be changed with a GFD. What about intel? They have to be able to be set externaly, or Intel would not be able to set the multiplyer when they &quot;bin&quot; them. SO, I am just saying, from my limited knowledge on the subject, there must be some way to override the onchip lock from intel.

Just my .02

 

[CTho9305]

someone here mentioned something about &quot;fuse locks&quot; which are &quot;blown&quot; and once they are the mult is locked. I dont remember who said it in which thread . BUT.... theoretically, it is possible that some of these &quot;locks&quot; which are not &quot;blown&quot; could be blown by the user, and that raises the mult (assuming all open is a higher mult). or, if the locatoin were external (anywhere outside of hte actual core) they should be mod-able