P45 DS3L quick OCing question.

[Koko56]

What should I do about the voltages that are set to Auto?

Would they increase proportionally with the FSB or just stay at their auto settings?

Also on the Asus P5B at least there was an option to set the frequency of PCE cards (33.33Mhz or something) but I have not seen anything like that on this motherboard.

Just a note:

CPU: E8400
RAM: 2x2GB Corsair XMS2 800Mhz C5

One other thing - CPU termination voltage. I've glanced at a few things, and got the impression that at that voltage the link between the CPU and the... NB(?) breaks/stops? Or am I completely wrong and at tht voltage the system would shut down as a safety feature?

 

[sirmonkey1985]

you should be fine leaving those on auto.. but if you want you can manually set them to the default voltages that show up to the left of auto setting.. as for the cpu termination voltage i dont know since i havent used an intel system since the P2 days..

 

[BillParrish]

The motherboard will set them to what is needed to boot as you change the FSB. How intelligently it does that is open to debate.

For anything other than very extreme OCing the only ones of concern and the only ones that should be other than default are.

Vcore
MCH core - set to +.1 for extra stability (see post below)
Dram

Leave the reference voltages and PPL on default. Typically they control the signal levels that determine what a valid 1 or 0 is on the data and address busses and so can screw stuff up quickly with very little benefit. In some very extreme cases very minor adjustments can help a tiny bit. You are nowhere near where you need to be to play with those and there should be no need for the motherboard to adjust them either. Play if you like, at worst the board will not boot and reset itself but they just add another variable in OCing you don't need right now.

anywho thats my 2 cents. and put a fan on the MCH/Northbridge if you go to +.2

 

[Varmint]

Vcore
MCH core - set to +.1 for extra stability and raise to +.2 if needed for high FSB OCing.
Dram

high FSB means relative to motherboard's limit or the CPUs? Like for my e5200, I can't hit 350 FSB, would raising MCH help? (it's already 1.2V).

 

[BillParrish]

The motherboards FSB limit. The MCH gets OCed exactly like a CPU when you raise the FSB. Just like the CPU it has to handle the higher speed of data and addresses on the buss and actually has a more difficult job as it has to deal with a high FSB on the buss from the CPU and also has to then manage and keep in timing passing that data to the memory (and the reverse) at what might be some different speed due to memory speed mulitpliers.

So like a cpu the MCH likes a little extra voltage when pushed hard and making sure the cooling is working as best it can (replace factory paste) or enhancing it (adding a fan) can not hurt at all.

For the P45 MCH the default standard voltage is 1.1V and per the Intel spec sheet the Absolute Max Vcc is 1.21 volts so I cannot recommend going much higher. I am going back and editing my previous post to reflect that, I was thinking about the P35 chipset or something with the +.2 comment.

However many people run the MCH way above Intels recommend Absolute Max Vcc and with good cooling there seems to be no ill effects. The Absolute Max Vcc is obtained by computer simulation and Intels knowledge of the typcial charistics of the manufacturing method and the type of technology used and is very conservative. Not on hard testing to failure of parts. So bottom line its your stuff and do what you need to do but I personally would not go over say 1.25V just because I dont want to spend money replacing a board to gain a few 100MHz. Other people have other goals and thats fine.
So I messed up in the post above and I am going to fix it.

Thanks for bringing that to my attention.

 

[Koko56]

Thank you guys!

Though what exactly is the FSB termination voltage?

Please.

 

[Biskquik]

From what I've read over at the Gigabyte support forums is that the PLL tends to overvolt on auto if overclocking. I have the same exact board but I'm using Q6600 running it at 3.0ghz (333*9). Below are my voltages:

CPU (Normal...Manual Setting)

CPU Vcore….………........1.28750V.................:1.28750V (I could undervolt further but I cringe when vDroop brings this down to 1.20)
CPU Termination…..... 1.200V.......: 1.2V
CPU PLL…………………....1.500V.......: 1.55V (I may try bringing this back down to stock)
CPU Reference.…………....0.805V........: Auto

MCH/ICH
MCH Core…………….....1.100V...........: 1.2V
MCH Reference….…….0.800V...........; Auto
ICH I/O……………….....1.550V............: Auto

DRAM
DRAM Voltage ……....1.800V............: 1.8V

Your termination voltage seems a bit high to me though.

Edit: found a thread about Termination over at another forum:
http://www.xtremesystems.org/forums/archive/index.php/t-103353.html

 

[Koko56]

I might now lower the FSB termination to about 1.2 and see what happens as well as putting the PLL to about 1.5V too!

Thanks for the tips!

EDIT: Yeah - the vDrop is quite big. I set the voltage to about 1.23 or something and I got 1.16 with an OC - BSOD! Then I just increased it and all was fine.

 

[BillParrish]

I think you will find putting the PPL voltage back to normal is a good idea.

The PPL voltage is the voltage given to the Phase Lock Loop clock generation chip. It has nothing to do with the voltage or frequency generated by the clock chip itself, it just raises the voltage supply to the clock chip. IMO its one of those "lets throw it in there to make the bios look cool" things. If you power supply is up to snuff and there is nothing wrong with the board there really should be no need to increase the voltage to the clock chip as they put out a very accurate and stable clock signal at stock voltages and increasing the voltage to the chip really should not affect its operation as the thing is designed and easily capable of providing clock frequencies way above what is needed. A one notch of extra voltage would not hurt, I just doubt it would help.

Here is the best discussion of FSB termination voltage (it is called Vtt in the article) I have found. Bit technical but understandable if you go slow. http://www.thetechrepository.com/showthread.php?t=87

Basically you are adjusting the voltage level at which the buss recognizes a logic one or zero in an attempt to compensate for the skew the data on the buss experiences at higher FSB. More often than not you fark things up. This is a very advanced tweak and should be left until one of the very last things you investigate in obtaining those last couple of MHz.

 

[Varmint]

My P45-UD3R seems to need 1.22V MCH do hit just 333 mHz FSB with my e5200.

Boots up fine at 300 mHz (x8), 1.1V, but won't boot at 333 (x8) FSB unless I raise the MCH.

That's why it seems like the CPU is a factor as well, since in theory the UD3R shouldn't need 1.2V to run such a low FSB, right?

 

[Koko56]

My P45-UD3R seems to need 1.22V MCH do hit just 333 mHz FSB with my e5200.

Boots up fine at 300 mHz (x8), 1.1V, but won't boot at 333 (x8) FSB unless I raise the MCH.

That's why it seems like the CPU is a factor as well, since in theory the UD3R shouldn't need 1.2V to run such a low FSB, right?

The CPU voltage in your sig does not appear to be the problem - so your logic relating to the MCH V might be correct, though my knowledge of how everything would relate is limited!

But I would imagine that a board like you've got would be alright up to 400FSB minimum.

BillParris - thanks for the explanation and that link. It's a bit too much for my head right now, but might come in useful at some time.

One other thing - I can't get FSB past about 385, I wonder what is limiting me here. I don't wanna go too high on the CPU voltage until the Zalman 9700 gets here.



Note: the voltage fluctuates from 1.232V to 1.216V when testing.

Oh yeah - I set the CPU termination to 1.2V and set the CPU PLL to 1.55 - the lowest given value.

 

[BillParrish]

It is likely time to up the Vcore as you are basically at stock CPU voltage according to that screen shot. While waiting for your new heatsink is wise and as there is not any hurry it is the smart thing to do.

However, if you find you cannot resist the impulse to "play" while waiting for the UPS truck to arrive some info for you.

Your CPU (very conservative numbers follow) is rated by Intel for a max Vcc of 1.45V and if you allow the .05V spec for cpu voltage regulation overshoot due to large load changes that results in a CPU voltage "ceiling" of around 1.4V so you have tons of room to up the voltage.

That CPU at around 100C (described incorrectly as Tjmax in Coretemp or RealTemp) your CPU will send out a signal to the motherboard that it is getting hot and request the motherboard to spin up all the fans to 100%. This is the PROCHOT interrupt and can be found in Intel documentation for your CPU. Several deg C above that, the exact number is undocumented but several hints in Intel docs indicate its 10C-15C higher, the CPU will just shut down to prevent damage. The point is that your screen shot shows you at 54C under full load. Your CPU is not even breaking a sweat.

So if you wanted to raise the Vcore CPU voltage to 1.35 and just see how that affected your ability to overclock it is extremely unlikely to cause any problems. You would be nowhere near any max temps or voltages. And of course you can always raise it up, see what happens and then set it back for the days use of the machine until the heatsink shows up.

 

[Koko56]

I was not itching to increase the voltage, but after that post I just might have to... or maybe just wait for the HSF. Dono.

As far as the temps go - yeah they look alright, and considering that's with the new, smaller stock HSF for the E8400, that is pretty damn good!

And of course - thanks for all the information that you people gave me!