Loadline calibration -- your thoughts?

[blanghorst]

OK, so I've read various pros and cons about enabling Loadline Calibration in a system's BIOS, particulary with 45nm chips. So I wanted to get everyone's thoughts on it. For what it is worth, I am using it to get a stable 1.36 V to my E8400 and it drastically improved stability. I'm concerned with the article on Anandtech about possible damage, though. So, I wanted to get everyone's thoughts in this thread and in the associated poll.

 

[BillParrish]

Are you speaking of this blurb ?

http://www.anandtech.com/mb/showdoc.aspx?i=3283&p=17

Loadline Calibration - [Enabled, Disabled] Setting Enabled will effectively remove any negative feedback in the CPU supply voltage regulation circuit. This means that there will be little to no difference between the CPU supply voltage at idle and full-load. This does not mean however that there will be no offset seen between the idle voltage and the CPU Voltage VID as selected in BIOS. Disabled follows the Intel specification and allows the CPU supply voltage to droop under load as properly designed. We recommend you leave this Disabled and increase CPU Voltage as needed to satisfy the minimum operating voltage needed under load.

My thoughts (take note of that, this is my opinion nothing else).

First understand exactly what and why. This is helpful.
http://www.thetechrepository.com/showthread.php?t=126
also the Intel specification for CPU Voltage Regulation Circuitry.
http://www.intel.com/design/processor/applnots/313214.htm
(section 2.2 and 2.7 are of specific interest. and figure 2.16 is the core of the matter)

Now having a background into what is going on, and why, no I would not be concerned with using load line leveling UNLESS I was within .075 volt of the CPUs Absolute Maximum Vcc when overshoot or other transients could cause the Absolute Maxium Vcc to be exceeded for more than 25uS.


Trouble is that the manufacturer of your motherboard, or any motherboard that I know of, provides NO assurance the VRD circuit will meet intel specs and NEVER have I seen true specificaions on the noise ripple current capacity or even voltage regulation of a boards VRD circuit. Just marketing crap about phases and solid caps and "ours is better" kinda bullshit for Joe Sixpacks.


So, if processors where not so cheap and I intended to run the cpu as absolutely hard as possible without ANY "comfort zone", very near, at or above Intels max Vcore, I would have an O-scope on the Vcc pins of the processor while rapidly changing the CPU load (and thus the current demand) and looking closely at the maximum value and duration of the voltage spikes produced by the cpu voltage requlation circuit in regards to the maximum CPU voltage specification.

Otherwise I would just stay under the max Vcc spec by .075 (a number I determined somewhat "offhand" by looking at table 2.14 which specifies a max overshoot of 50mV and giving myself an additional 50% safety facor of + 25mV resulting in a total of .075V . )

As mentioned above there is NO assurance that your or any motherboard complies with the Intel spec of 50mV overshoot or even 75mv.

Those are my thoughts, I accept no responsibility for anything and you cant make me.

 

[blanghorst]

Yes, that is one of the articles I was referring to in my original post. Anandtech (and others) advocate moving your voltage up in order to get the desired Vcore rather than using LLC and having a lower voltage setting in the BIOS.

For example, on my board, I have the Vcore set to 1.37 or so in the BIOS. With LLC enabled, this results in a stable 1.36 volts to the CPU as measured by CPU-Z. If I disable LLC, it drops to 1.34 and of course, would drop a little further with load. If we assume that it is a .03 V Vdrop from the example above, Anandtech would say disable LLC and make the Vcore in the BIOS 1.39 V if you need 1.36 V for stability.

I agree with what you state though and it is along my same lines of thinking. I am getting a solid 1.36 V now. Even with the .06 or .07 volt transient spikes, that is well below the 1.45 V "max voltage" of the E8400, so I think I am safe. I was just wanting to see if my thinking was wrong on this or if I missed something.

 

[ultimate11]

my q9300 at stock when idle is crashing , its stable on full and partial load but when i go sleep and get up the bsod is there , i tried upping the nb to 1.4 but no luck and have LLC disabled . Is there anything else that i must do to get a stable result?

 

[BillParrish]

it is either heat or software. Write down exactly what the BSOD says and why the hell are you letting the machine run all night doing nothing ? I doubt it is at idle all night, you must be downloading something, folding, chinese hackers have "botted" you and 120,000 unauthorized emails later the machine crashes, something.

What are your core temps at idle and load ?
What does the BSOD say EXACTLY ?
What is the machine doing all night ?

You should have made a new thread. Perhaps add the info above and repost.

 

[ultimate11]

i just have downloads going on , i stress test it all day and play games like grid/aoc all day long

my cpu cores are idle around 44-46 and at load they are 50-52 max (cpu temp is 32-36 at all times)

bsod is not giving any error about .dll or .sys there is a problem w/ rampage and LLC and after googling

up most of ppl are having the same problem which they are resolving by disabling LLC and upping the

vcore and nb. However i cannot find cpu Vcore in the bios (does it goes by another name?) so i can set

everything up tonight and make it go for a test run.

i also memtest the ram its all good

this is the link i read http://www.xtremesystems.org/forums/showthread.php?t=189507&highlight=asus+rampage+formula

ps : i am not ocing at all , my extremetweaker settings are on manual but fsb and multi are on default (so i can turn off LLC like that)