stn0092

I'm doing 470x9 on my Q9650 which will only pass small FFTs with 1.46XXv in BIOS w/ LLC, 1.440 load and idle. My vdroop without LLC is even worse than your 0.100v. I can boot with vcore as low as 1.40 but it'll BSOD the moment I begin a stress test; values in between will do the same or fail tests.

I've seen plenty of people ramping up to 1.50v vcore in the UD3P thread at XS so, compared to thenm, 1.45v is acceptable to me.

Yomny

not speaking from experience because im not expert but i really dont know why LLC is not suggested with as it seems to only help with the actual and under load values of the voltages. Speaking with only my common sense and a bit of knowledge i say try the LLC and lower the actual bios voltages and get a safe windows vcore then the under load vcore should be closer to the actual vcore displayed in windows.

Robilar

I expect Eric will chime in with the technical explanation on LLC but in a nutshell, at idle, LLC causes severe voltage spikes. The older 65nm chips are far less voltage sensitive and thus are unaffected.

The 45 nm chips can be damaged by this effect. I fried an E8400 after about two weeks of running a moderate overclock on an asus board using LLC.

In my experience, the guys that are clear about it being dangerous tend to be the most experienced benchers at XS. Fellows like Fugger stay far away from it and he cooks hardware for giggles.

I agree that in theory LLC can provide much lower required vcore under load, but idle seems to be where the issue lies.

arjo_reich

See, I keep asking if I should believe what OCCT is reading or if just caulk it up to the fact that I tend to run Prime95 & OCCT simultaneously... seriously though, look at those spikes. Several people told me those are bunk but at least as many other people refer to the fact that LLC causes these very same types of problems...

Yomny

thanks a lot for the explanation Robilar, what i didn't know was that LLC cause severe spikes in voltages at idle, and it makes sense why more voltage sensitive chips could be harmed.

colin niloc

Would HW monitor pick up the cpu vcore spikes people are talking about if you have LLC enabled?

I have ran HW monitor for since 12-12-08 without restarting my rig and the cpu vcore has not changed from 1.39v it reads 1.392v in cpu-z

__________________
Lead me, follow me, or get out of my way.
-G.P.

arjo_reich

it's not much @idle as it is at the precise moment in which a core is about to go underload and the vCore for the processor is about to experience vdroop, LLC kicks in and regulates the voltage very quickly.

To put an analogy to it, it's almost like driving in rush hour and just as soon as you realize that someone is going to merge in you floor it to close in the gap... well, most of the time that's ok but at some point the person in front of you is going end up breaking at the same time and you'll get a collision.

The LLC regulates the voltage by overcompensating for the vdroop but because it doesn't know how much or how long the drop is going to be for, it has to overshoot it some before returning to nominal. If you're already close to your chip's frying point, such a spike could be devistating - especially for 45nm chipsets.

:shrug:

colin niloc

Pretty good analogy. Maybe this board does better then other boards at regulating the voltage or what ever it does. I have ran prime 95, encoded movies, played CS:S and a bunch of others stuff without any change in vcore according to HW Monitor. I just opened up cpu-z started prime 95 while typing this still no fluctuation in the core voltage on either.

__________________
Lead me, follow me, or get out of my way.
-G.P.

colin niloc

Here is a screen shot

Attached Thumbnails

 

__________________
Lead me, follow me, or get out of my way.
-G.P.

stn0092

So it's better than I crank up vcore to 1.55 in BIOS to get the 1.44 I want under load?