xtremeoverlocking - pushing the 3930K/3960X to 4.5 GHz - update

[BeepBeep2]

First off, you seem to be the only person that's raising a stink about it. Second, it isn't about a who's-blaming-who competition (because if that were the case, you'd win by a LONG shot). Third:

quote:
"Base clock adjustment is rather limited, the reason why you bought a K processor with unlocked multiplier."

No. "We" bought this K processor because it's cheaper than the X-series. I don't think that you can even GET a 6-core Sandy Bridge with a locked multiplier anyways, so...that's nullifies your point.
Okay? You also bought it because it has an unlocked multiplier, otherwise you couldn't overclock it much.


You DO realize that by using acronyms without first defining what the acronym is, you're making an assumption that the target audience already knows what you're talking about. (Clearly, as you've already said and as I've indicated, I'm NOT a regular overclocker, so WHYYY would you assume that I already know that???) It is for the same reason why I will ALWAYS expand an acronym at least ONCE before using said acronym from that point forward. Now, whether you actually understand what comptuational fluid dynamics (CFD) is or finite element analysis (FEA) is - that's relevant only to the point of "this is what I'm testing", but given that this isn't a CFD or FEA course, that's another discussion for another time. And to your point, even finding out what the acronym is - doesn't actually explain it as you have. (I'm downloading the manual now to check/cross-reference that since you didn't include a page reference/citation that the manual actually explains what load line control is, but I'm going to assume that you're telling me that it's in my manual because it actually explains it in the manual as you have. And if that WASN'T your point, in that LLC only exists in the manual as the expansion of the acronym, you could just have easily pointed me to acronymfinder.com for the same purpose. At which point, at least I would know what the acronym is, but not necessarily what it means or what it does (much like how I have expanded CFD and FEA, but probably doesn't tell you much about what it is or what it does).

Who the **** cares? You obviously knew what acronym I was talking about. What else would have have been talking about? We were discussing Load Line Calibration/Control. If you've spent more than five minutes on any technology enthusiast forum like this one you'd realize that acronym use is common and people know or figure out what they mean quickly without trying to insult the person using them...

*edit*

actually....LLC does NOT appear in the manual at all. Do a search for it. (Disproved.) Now, had I known that it LLC is load line control, and that's how I know to search for say..."load line", THEN it shows up (first instance at p. 44). But that requires knowledge beforehand.
Who cares? You already DID have that knowledge beforehand...you demonstrated that here:
http://www.xtremesystems.org/forums/...=1#post5090230

Ahhh...so that's what that is. I dunno. I COULD be wrong, but I think that LLC in that case/sense then has been in workstation/server boards for a really long time then.
...
P.S. Where's Vdroop on my BIOS? I don't even see it. I think that I do remember seeing an option for load-line calibration though, but I forget where it is on the GA-X79-UD3 with an F10 BIOS.

And for computers in general (not necessarily limited to CPUs, though I suppose the concept could be extended to CPU crossbar switches); LLC could also be an acronym for logical link control.

See: http://en.wikipedia.org/wiki/Logical_link_control

(Course, I'm going to bet that you're going to say something along the lines of "this is just more random that's irrelevant, except that it illustrates how LLC can be referring to other things and the inherent presumption that just cuz YOU know what it's loadline, does NOT necessarily indicate or predicate that other people do. ) --- which I'm sure now you're going to call THAT "nitpicking". You're like a woman (or an environmentalist) - you can never win with them. If it's not clean air, it's clean water. But it's always gotta be SOMETHING that's wrong.
STFU, you already knew what "LLC" meant within our discussion but chose to pick a fight with me for no reason other than to give yourself satisfaction later in the thread. You seem to be the "woman" here...creating situations out of nothing significant so you can win. It's always gotta be "SOMETHING that's wrong." in whatever I say to you.


From Anandtech's article on the 3960X:

"With a bit of effort I had no problems hitting 4.6GHz on my Core i7 3960X review sample. I had to increase core voltage from 1.104V to 1.44V, but the system was stable."

Source: http://www.anandtech.com/show/5091/i...gh-end-alive/3

So per that - that would seem like that my 1.42 core voltage is right in line with that.
Okay? But your voltage during stability testing is ~1.37v?


I don't know about you, but have you ever noticed that whenever people say that they want someone else to stop doing something, they're often the culprits of the identical or similiar offense? Think about it. Or just sit and observe.

Re-read both threads in case you need a reminder of the manner of your responses.

In fact, from the initial thread - do you want to know what's the FIRST thing you said to me?

Quote:
"Why come here if you already have your mind up? You seem to be talking the talk, a bit over your head, if you know what I mean."

So, who's making it personal and "nitpicking/insulting" who now first?
Well, you came in here acting like you already knew everything...

Secondly, my initial response to that did even use the pronoun "you" (thus making it personal a la vous):

"Because that's based on my initial research. Obviously, if ther'es a better solution out there, I'll go with that. (I ended up pricing the system based on what Leeghoofd suggested, so how's THAT for "why come here if you already have your mind up?"??? :P"

So, once again who's insulting who now?

*edit* Damn...I am what I hate because I'm taking random puck shots now. (Too easy.)
I'm glad you realize...you have been for a while now.


Well...1.36 V vs. 1.42 V - that's 4.2% difference. Normally, it wouldn't matter so much, but if/when 0.01 V can result in a run going from unstable to stable -- 0.01 V out of a 0.06 V difference -- that's 17%!!! that's HUGE!

And I fully agree with you - 1.40 is not equal to 1.42, but at least that's closer to 1.42 than 1.36. That and the fact that you can't really set the Vcore @ load...or at least it doesn't appear that way. (I don't remember what my loadline control settings are - but it's probably whatever the defaults are.)
Yes...
Honestly, if it were me, and if I really was concerned about it; I'd be asking about how DOES a SOFTWARE measure Vcore? Now, if someone were to come back and say that there are sensors built into the IC, then why would different programs end up with different values. CoreTemp not equal to CPU-Z not equal to HWMonitor even if I were to suppose that they are reading from the same sensor chip? That to me, just seems suspect. Three programs, and I can't get a single answer - so something's not jiving.
I told you already, here:
QUOTE=BeepBeep2;5090189]You can talk all you want, but you are misunderstanding things here...

CoreTemp is reading CPU-VID. That is the voltage that the CPU is set by intel to run at for either the default or Turbo pstate. If you do a text dump from the CPU-Z validation tab and open it, you will see all the VIDs are for each pstate.

(P.S. In the last three years, I'm certain that SOMEONE has certified/validated the readouts from the programs somehow. I'm not sure how exactly you'd tap into the IC between the CPU and the socket to actually measure that, but let's suppose that they actually found a way to directly measure it, I'm CERTAIN that in three years, someone's been able to prove that the results from one of those three programs is statistically accurate 19 times out of 20 with some kind of sampling rate error. And that they're testing with some kind of ASTM or ISO or IEEE method or if they have their own criticisms of "standard test methods", they've been able to identify what they are and why they are and their modified versions of it with their own logic and reasoning for why they think the test should be conducted in a different manner (and they might even have written a paper illustrating how their modified method is more accurate than the standard method).
...didn't I mention measuring with a Digital Multimeter before in this thread? The software isn't always accurate, sometimes not even close. Common measuring points are certain IC legs (can be looked up in whitepapers) or on the legs of capacitors/resistors. The software readings are there to give a general idea, and there are inaccuracies within even different boards of the same model. Nothing is perfect, and manufacturers don't/can't spend time calibrating these readings.
I would LOVE to read those papers if you've got 'em. Can't necessarily guarantee you that I'd actually UNDERSTAND any of it, but...at least I'll have a look-over. And try to understand why three separate programs can result in three different value. Or perhaps what's more interesting, both CoreTemp and CPU-Z read out to 5 significant digits while HWMonitor only read out to 3 significant digits. Hmmm...(and it doesn't look like that you can set it either). And by the rules of significant digits, CoreTemp and CPU-Z would be more precise, but not necessarily any more accurate.
Yes, but I already explained to you that CoreTemp is not reading the actual running voltage while HWMonitor and CPU-Z are...HWMonitor and CPU-Z are made by the same developer, and read the same voltages...but it doesn't even matter because I already told you that sensor readouts are in .012v steps negating the need for 4 significant digits.

You want to know what the manual says in regards to CPU Vcore?

"Core Voltage Control``
This section provides CPU voltage control options." (p. 46)

That's IT. It says NOTHING about SETTING it in the manual. (so...there goes THAT theory.) So contrary to SOMEONE'S belief, yes, I DID read the manual.
Great!
And as indicated, I was googling Vtt and PLL to try to understand what they do.

So there goes THAT theory. What else you got?
Great!...I don't really care if Gigabyte spent five seconds on their manual...
I AM asking you more questions. But then you'd call them as either nitpicking or irrelevant or you'd skim RIGHT over them. So...how could you dare say "ask more questions instead of calling me stupid" when I am ALREADY doing JUST that.

Heck, in fact, I think that after taking the CPU off auto Vcore, I was using 1.30, 1.32, 1.35, 1.38, 1.39, 1.40, and then finally 1.42 in that sequence, in that order to try and get the system stable.

From 1.30-1.38; that initial testing was done with IBT. Once IBT was stable at 1.38, that's when I started with the crash sims and it had further stability issues, so from 1.39 to 1.42 - that was the sequence taken to get the crash sim to be stable. And no, I didn't try 1.41.

From 1.30 to 1.38 - that was directly from the voltages that people were reporting/suggested range for me to try and so I started with that. So, again, contrary to your belief, the information that is presented here IS put to use and not ignored. (Yes, I AM assuming or reading that that's what your belief is based on the premise of your above statement.)

Now, if that's the Vcore@load, and given that we've already established that the Vcore@load does fluctuate with different programs and different loads, I would have had to calibrate that in with the program that they were using in order to get an agreement and getting the system to jive and the numbers to talk to each other. But that should also mean that along with that, a separate value should be presented as "this is what I set in the BIOS". Because you can't really control what the Vcore@load is directly. Nor am I aware of any method/procedure for predicting what the Vcore@load given x load is going to be. So...you work with what you've got - which is namely, the Vcore setting in the BIOS (@idle) and loadline control. (Which I'm sure YOU know, but for someone like me who's new to this - I didn't. Until now that is.)
It really isn't that big of a deal...this isn't rocket science.

And it is PRECISELY for this reason why I NEVER assume people know what I know unless they indicate otherwise. (The most telling sign for me typically is by the questions that they ask as a result of it.) Because if you assume people to know what you know, you run into this -- EXACTLY this type of a situation where you thought someone else knows what you know, the documentation didn't adequately cover the material, and you write someone off as asking idiotic questions (which you then considered irrelevant, nitpicking, idiotic, or you just skim right over them as if they weren't important. Well if they weren't important, why would someone be asking them in the first place??? Especially when that someone's already stated that that someone RARELY asks questions for the benefit of their own health or rhetoric?)
Okay, then don't assume...ask them nicely instead of saying things like "I can't help it if you can't xxxxxxxxx."[/QUOTE]

I got tired of splitting quotes. I don't even want to answer any more of these, I have a life to get to.

[Leeghoofd]

UD3 will be just fine at 4500mhz... no need to cash out again... Just get this one stable and forget about it... Maybe you just need a beefier cooler, plus make sure the top of the board gets extra ventilation ! Aim for 1.45 in bios, will droop/drop to 1.4ish under load...

Chill Beep man... chill

[BeepBeep2]

UD3 will be just fine at 4500mhz... no need to cash out again... Just get this one stable and forget about it... Maybe you just need a beefier cooler, plus make sure the top of the board gets extra ventilation ! Aim for 1.45 in bios, will droop/drop to 1.4ish under load...

Chill Beep man... chill

Uhh, he knew what LLC meant but then tried to make me look stupid because "LLC" wasn't in the manual I said it was in, only "Load-Line Control"?
...and then said pretty much "Had I known, or you had been more clear...you could have meant unrelated thing X, Y, or Z,...and I didn't know..."

Then he forgot/didn't read what I said about CoreTemp but tried to use that against me too...

But he only did it to nitpick, to pick a fight, to start an argument...this guy doesn't really deserve any advice.

[alpha754293]

Uhh, he knew what LLC meant but then tried to make me look stupid because "LLC" wasn't in the manual I said it was in, only "Load-Line Control"?
...and then said pretty much "Had I known, or you had been more clear...you could have meant unrelated thing X, Y, or Z,...and I didn't know..."

Then he forgot/didn't read what I said about CoreTemp but tried to use that against me too...

But he only did it to nitpick, to pick a fight, to start an argument...this guy doesn't really deserve any advice.

Dude! I ALREADY TOLD YOU THAT I DIDN'T KNOW WHAT IT MEANT BEFORE.

What's LLC? and what's VCCSA?

"Yesterday 02:08 PM"

http://www.xtremesystems.org/forums/...=1#post5089975

For 4.5ghz, auto settings are well within a safe range. If you wanted to drop idle vcore a little you could try raising LLC/Load Line Calibration.

http://www.xtremesystems.org/forums/...=1#post5089995

"Yesterday 02:56 PM"

2:56 PM > 2:08 PM

Eff man. What the HELL is wrong with you?

[OC Nub]

Come on guys, lifes too short to keep this going.

BeepBeep2 is a very good over clocker. If he is willing to offer help, I would take it and just say "Thanks".

[alpha754293]

First off, you seem to be the only person that's raising a stink about it. Second, it isn't about a who's-blaming-who competition (because if that were the case, you'd win by a LONG shot). Third:

quote:
"Base clock adjustment is rather limited, the reason why you bought a K processor with unlocked multiplier."

No. "We" bought this K processor because it's cheaper than the X-series. I don't think that you can even GET a 6-core Sandy Bridge with a locked multiplier anyways, so...that's nullifies your point.

Okay? You also bought it because it has an unlocked multiplier, otherwise you couldn't overclock it much.

ALL Sandy Bridge 6 cores are ALL unlocked.

You DO realize that by using acronyms without first defining what the acronym is, you're making an assumption that the target audience already knows what you're talking about. (Clearly, as you've already said and as I've indicated, I'm NOT a regular overclocker, so WHYYY would you assume that I already know that???) It is for the same reason why I will ALWAYS expand an acronym at least ONCE before using said acronym from that point forward. Now, whether you actually understand what comptuational fluid dynamics (CFD) is or finite element analysis (FEA) is - that's relevant only to the point of "this is what I'm testing", but given that this isn't a CFD or FEA course, that's another discussion for another time. And to your point, even finding out what the acronym is - doesn't actually explain it as you have. (I'm downloading the manual now to check/cross-reference that since you didn't include a page reference/citation that the manual actually explains what load line control is, but I'm going to assume that you're telling me that it's in my manual because it actually explains it in the manual as you have. And if that WASN'T your point, in that LLC only exists in the manual as the expansion of the acronym, you could just have easily pointed me to acronymfinder.com for the same purpose. At which point, at least I would know what the acronym is, but not necessarily what it means or what it does (much like how I have expanded CFD and FEA, but probably doesn't tell you much about what it is or what it does).

Who the **** cares? You obviously knew what acronym I was talking about. What else would have have been talking about? We were discussing Load Line Calibration/Control. If you've spent more than five minutes on any technology enthusiast forum like this one you'd realize that acronym use is common and people know or figure out what they mean quickly without trying to insult the person using them...

*edit*

actually....LLC does NOT appear in the manual at all. Do a search for it. (Disproved.) Now, had I known that it LLC is load line control, and that's how I know to search for say..."load line", THEN it shows up (first instance at p. 44). But that requires knowledge beforehand.
Who cares? You already DID have that knowledge beforehand...you demonstrated that here:
http://www.xtremesystems.org/forums/...=1#post5090230

Ahhh...so that's what that is. I dunno. I COULD be wrong, but I think that LLC in that case/sense then has been in workstation/server boards for a really long time then.
...
P.S. Where's Vdroop on my BIOS? I don't even see it. I think that I do remember seeing an option for load-line calibration though, but I forget where it is on the GA-X79-UD3 with an F10 BIOS.

Load line calibration is actually NOT listed in my Tyan S2915WA2NRF manual. My apologies. I misspoke. I was thinking spread spectrum (which I'm guessing is something totally different).

STFU, you already knew what "LLC" meant within our discussion but chose to pick a fight with me for no reason other than to give yourself satisfaction later in the thread. You seem to be the "woman" here...creating situations out of nothing significant so you can win. It's always gotta be "SOMETHING that's wrong." in whatever I say to you.

No, because it's called nitpicking, or irrelevant, or whatever the hell you're going to call it next.

Okay? But your voltage during stability testing is ~1.37v?

Did I SET it at 1.37 V? No. What DID I set it at? 1.42 V. So what's your point?

Even from Anandtech:

"I HAD to increase core voltage from 1.104V to 1.44V."

I don't geddit. What's so hard to understand about that? Notice how he doesn't report what the stable voltage is at load. In fact, in that whole page, that's the ONLY place where the word "voltage" shows up.

Well, you came in here acting like you already knew everything...

If I knew, I wouldn't be asking.

(Notice that I didn't ask you guys how to run crash sims. Geee...I wonder why!) (Granted, I'm no expert in that either...I'm still learning.)

I'm glad you realize...you have been for a while now.

LOL...

Sheldon: "That was sarcasm?"
Me: *touches nose*
Sheldon: "Yes! I'm on a roll!"

& *end scene*

I told you already, here:
QUOTE=BeepBeep2;5090189]You can talk all you want, but you are misunderstanding things here...

CoreTemp is reading CPU-VID. That is the voltage that the CPU is set by intel to run at for either the default or Turbo pstate. If you do a text dump from the CPU-Z validation tab and open it, you will see all the VIDs are for each pstate.

This is a perfect example of what I had called an "incomplete" answer before.

Okay, so you told me what CoreTemp is reading.

Now, I'm going to assume that CPU-Z is also reading the same thing (namely CPU-VID) by merit of CPU-Z's validation tab and dumping out the text file which shows the VIDs for each Pstate. Fine. BUT that STILL doesn't explain why you've got two programs (CoreTemp and CPU-Z), BOTH reading CPU-VID, and coming up with different results. Especially given that they're reading the SAME CPU-VID. Now why is that? Why would two programs reading the SAME CPU-VID give me two different results?

Does HWMonitor read the same CPU-VID? (I'm not sure - I don't remember if you mentioned that it reads the same thing - because IIRC, you just said that most people here use that, but don't remember if you specific the method that (or source) that it gets its numbers from. (And if it IS also from CPU-VID, then my initial question still stands (which you would have now completely FAILED to answer - I've got three programs reading from the same source of data and I get three different answers. Why is that?)

...didn't I mention measuring with a Digital Multimeter before in this thread? The software isn't always accurate, sometimes not even close. Common measuring points are certain IC legs (can be looked up in whitepapers) or on the legs of capacitors/resistors. The software readings are there to give a general idea, and there are inaccuracies within even different boards of the same model. Nothing is perfect, and manufacturers don't/can't spend time calibrating these readings.

I don't have a DMM (if that wasn't obviously apparent). Now, if you can tell me where I might be able to get one on a Sunday night at 10:09 PM, please...by all means - enlighten me.

Yes, but I already explained to you that CoreTemp is not reading the actual running voltage while HWMonitor and CPU-Z are...HWMonitor and CPU-Z are made by the same developer, and read the same voltages...but it doesn't even matter because I already told you that sensor readouts are in .012v steps negating the need for 4 significant digits.

Wait...I'm confused now. Okay, so CoreTemp is reading CPU-VID.

If I do a text dump from CPU-Z validation tab, it writes out the VIDs for each of the Pstates.

So...how DOES CPU-Z read the voltages? Huhhh??? I'm confused.

[OC Nub]

Dude! I ALREADY TOLD YOU THAT I DIDN'T KNOW WHAT IT MEANT BEFORE.



"Yesterday 02:08 PM"

http://www.xtremesystems.org/forums/...=1#post5089975




http://www.xtremesystems.org/forums/...=1#post5089995

"Yesterday 02:56 PM"

2:56 PM > 2:08 PM

Eff man. What the HELL is wrong with you?

Sorry man, no offense, but I was giving you way too much credit. I assumed you would know to lower vcore before raising LLC.

You screen shot showed 1.388vcore load in CPUZ, and you have 1.42v set in the bios.

In order to reduce the amount of voltage drop/droop you would raise the level of LLC but also lower the vcore. Ideally what most people want is to have idle and load vcore to read the same. Personally I dont use high LLC, but thats just my preference.

If you know your CPU requires 1.38vcore load, and you have 1.42v set in bios, then you can just start lowering vcore, and raising LLC, fire up your stress test or program you have been using, and look to see what your load vcore is and adjust accordingly.

The best way to learn how to over clock is to start at the beginning. Not shoot for the moon right out of the gate.

If you somehow killed your chip by my advice I will send you another 3930k. I shouldnt have assumed you knew more than you do.

[alpha754293]

Come on guys, lifes too short to keep this going.

BeepBeep2 is a very good over clocker. If he is willing to offer help, I would take it and just say "Thanks".

I'd have to sift through all of his posts to sort/find/filter out what's actually been useful vs. his b!tching.

Off the top of my head, I don't think that I've used ANY of the stuff that he's written in his posts cuz it APPEARS to be 99% b!tching.

I've used your info to leave VTT, IMC, and PLL voltages alone. I didn't get a chance to lower the PCH voltage.

I also used s4pphire's recommendations for the voltages from the other thread "Hitting 4.5 on a 3930 or 3960 will take roughly 1.3-1.38V". And that was my starting point.

And then I started upping the CPU ratio first, ran IBT, and until it became unstable, and then took Vcore off auto and put in the initial value of 1.30, and then progressively upped both CPU ratio and Vcore in that manner. (I think that it was like 32, 35, 40, 45 - in that order.)

And those were my starting points/starting inputs to this OC.

The term "offset voltage" not found in BIOS so I didn't do anything with it. Couldn't find "Vdroop" in the BIOS, so didn't do anything with that either.

[alpha754293]

Sorry man, no offense, but I was giving you way too much credit. I assumed you would know to lower vcore before raising LLC.

You screen shot showed 1.388vcore load in CPUZ, and you have 1.42v set in the bios.

In order to reduce the amount of voltage drop/droop you would raise the level of LLC but also lower the vcore. Ideally what most people want is to have idle and load vcore to read the same. Personally I dont use high LLC, but thats just my preference.

If you know your CPU requires 1.38vcore load, and you have 1.42v set in bios, then you can just start lowering vcore, and raising LLC, fire up your stress test or program you have been using, and look to see what your load vcore is and adjust accordingly.

The best way to learn how to over clock is to start at the beginning. Not shoot for the moon right out of the gate.

If you somehow killed your chip by my advice I will send you another 3930k. I shouldnt have assumed you knew more than you do.

You know it's bad when I make a reference to acronymfinder.com to try and find out what what LLC stands for. LOL...

(And yes, I actually did have to look it up on there to find out. I missed that you had actually mentioned before. Course, to his credit, after BeepBeep2 explained what it actually was to me - that make a lot more sense as to why the voltages in the screenshot was lower than what I had set in the BIOS. Didn't really think too much of it before. Course, I've also NEVER paid any where NEARLY this much attention to voltages (running or otherwise). so.... mehhh....)

[alpha754293]

Sorry man, no offense, but I was giving you way too much credit. I assumed you would know to lower vcore before raising LLC.

You screen shot showed 1.388vcore load in CPUZ, and you have 1.42v set in the bios.

In order to reduce the amount of voltage drop/droop you would raise the level of LLC but also lower the vcore. Ideally what most people want is to have idle and load vcore to read the same. Personally I dont use high LLC, but thats just my preference.

If you know your CPU requires 1.38vcore load, and you have 1.42v set in bios, then you can just start lowering vcore, and raising LLC, fire up your stress test or program you have been using, and look to see what your load vcore is and adjust accordingly.

The best way to learn how to over clock is to start at the beginning. Not shoot for the moon right out of the gate.

If you somehow killed your chip by my advice I will send you another 3930k. I shouldnt have assumed you knew more than you do.

Once again, I'm confused.

The manual says this about Vcore Loadline Calibration:

"Allows you to configure Vcore Load-Line Calibration. This item allows you to adjust the V-droop value. The higher the V-droop value, the lower the output voltage will be under heavy load."

And the allowable range is between 0.00 milli-Ohms to 6.00 milli-Ohms.

So, would I want to want to set that to 0.00 milli-Ohms or set it to 6.00 milli-Ohms? If they're saying that the higher this value is going to be, the higher the drop under heavy load, so that means that I should set it to 0.00 milli-Ohms, right?

What's the downside of doing so? (I'm guessing that there are pros and cons like anything/everything else???)

[OC Nub]

Here is an example, your board will have different names for the levels of LLC, but it should work the same.

LLC medium, turbo evo show vcore set in bios, and cpuz shows actual vcore load (or close enough without using dmm) , then idle.

[OC Nub]

Lowered vcore and raised LLC to ultra. Again first screen shot shows load vcore in cpuz, and turbo evo shows vcore set in bios.
Looks like Asus nailed this one, almost perfect with ultra high llc, load, idle, and what was set in the bios.

[alpha754293]

Try these , runs fine on my 3820 on the UD7 board with 16Gb 2133 ram. CPU here runs on air 4500mhz with Prime95 Custom set at 1.34Vcore loaded. (Bios 1.38) No need to alter voltages, CPU reacted similar on the UD3...

Attachment 126014

Attachment 126015

Attachment 126016

Attachment 126017

Attachment 126018

Attachment 126019

I just tired the settings that you have attached in your screenshots and the system keeps going on and off...repeatedly.

*edit*
The POST screen doesn't come up at all. How long should I let it keep going this on-off thing before I should clear the CMOS?

*edit*
It's been doing this for like 5 minutes now. Clearing CMOS...standby...

[alpha754293]

Lowered vcore and raised LLC to ultra. Again first screen shot shows load vcore in cpuz, and turbo evo shows vcore set in bios.
Looks like Asus nailed this one, almost perfect with ultra high llc, load, idle, and what was set in the bios.

Cool. So here's another question then -- why don't CPU/board manufacturers make it like that to begin with then? I mean, I'm guessing that having a lower Vdroop is MUCH better - so why don't they just make 'em that way always?

[OC Nub]

Intel designed the CPU's to this way, manufacturers have to design the board to be able to meet Intel specs (or something like that).

Im sure someone else can explain this better.

[alpha754293]

Intel designed the CPU's to this way, manufacturers have to design the board to be able to meet Intel specs (or something like that).

Im sure someone else can explain this better.

Hmmm...interrresting. I wonder why...I'm pretty sure that there's got to be a reason why they would do something a particular way. At least most of the times anyways...

[BeepBeep2]

Dude! I ALREADY TOLD YOU THAT I DIDN'T KNOW WHAT IT MEANT BEFORE.



"Yesterday 02:08 PM"

http://www.xtremesystems.org/forums/...=1#post5089975




http://www.xtremesystems.org/forums/...=1#post5089995

"Yesterday 02:56 PM"

2:56 PM > 2:08 PM

Eff man. What the HELL is wrong with you?

"Yesterday 11:22 PM"

Ahhh...so that's what that is. I dunno. I COULD be wrong, but I think that LLC in that case/sense then has been in workstation/server boards for a really long time then. Probably for the very same reason (at least from the load/unload perspective) although for workstation/server systems, they're probably wanting to squish out fluctations for longetivity reasons. And considering that my Tyan quad Socket 940 Opteron system is still handling the bulk of the simulation work right now still - being a system that's 5 years old, it seems to make sense that they would implement that or something like that much earlier on.

You abbreviated it yourself. You're gonna ask what the hell is wrong with ME? Also, I never told you that "LLC" was listed in your Tyan manual, I was suggesting it was in your Gigabyte board manual, and it was, in non abbreviated form.

Also, Anandtech (which you quoted) I believe was reporting idle voltages...(1.104 matching CPU-Z screen)...it wasn't exactly what they set, and it is possible that their load voltage was similar but you don't know. Therefore, you picked a bad data point.

You also say that my answer regarding CoreTemp was incomplete, but there was nothing incomplete about it. If you didn't know what I meant by "VID" or "P-State" then you should have asked instead of neglecting to take in the information. You can google that **** anyway but you obviously haven't.

On the CPU tab of CPU-Z, it is reading the same thing HWMonitor is. CoreTemp, like I said is reading the CPU VID for either the DEFAULT or TURBO P-State. That means either the 3.2 or 3.8 GHz P-State. It only reads one. Okay? Good. When you do a .txt diagnostic dump from the VALIDATION tab, within the text document will be ALL OF THE CPU P-STATES INCLUDING THE POWER SAVING IDLE ONES WAY BELOW 3.2 GHz.

I already told you this. Don't tell me some like "well you didn't tell me there were more than 2 P-States or what they were", take that upon yourself to type into Google and read for a few seconds.

EDIT: Actually, I believe I explained to you very clearly before that a VID is

the voltage that the CPU is set by intel to run at for either the default or Turbo pstate

and absolutely NOT a readout from the motherboard sensors in realtime, which I explained already was how HWMonitor and CPU-Z read voltages, and we discussed in detail.

I'm tired of you trying to nitpick

nit·pick (ntpk)
intr.v. nit·picked, nit·pick·ing, nit·picks
To be concerned with or find fault with insignificant details.

that doesn't matter or you don't understand because you can't read.

[alpha754293]

"Yesterday 11:22 PM"

You abbreviated it yourself. You're gonna ask what the hell is wrong with ME?

As I recall, that was AFTER you had explained it to me, was it not?

http://www.xtremesystems.org/forums/...=1#post5090219

Yes, that is what you told the motherboard to set. However, VR (voltage regulator) modules have a transient response to load and furthermore are not perfect...
There is a droop in voltage to the CPU as load increases on the motherboard VRM as well. This is part of intel specification, it is called "vdroop"...

Us overclockers hate vdroop, because when pushing overclocks, this vdroop can become quite large and make us set ridiculous "idle voltages" (what we call the voltage reading in CPU-Z/HWMonitor or on a Digital Multimeter at idle) that are possibly unsafe or make us flat out feel uncomfortable. In the past, we solved this by modifying power circuits by hand and soldering/changing resistors.

Within the last few years, motherboard manufacturers have implemented a simpler solution for us and made it easier (by request, of course!) by implementing "Load-Line Calibration" in BIOS so we can have equal load and idle voltages. Of course, the circuit still mantains a transient overshoot (can be quite large), but none of us really care as it is very short and/or has little to no chance of damaging our processors.

"Yesterday 10:58 PM"

Furthermore...

When we overclock on liquid nitrogen cooling at high voltage and frequencies (nearing 6 GHz and above) most of us set more aggressive LLC because vdroop becomes extremely large under the massive loads a 6 GHz CPU at high voltage can pull.

From the same post.

Feel free to the correct my logic, but 11:22 PM is > 10:58 PM is it not???

(Christ! You make it so easy!)

You abbreviated it yourself.

....after you...

You're gonna ask what the hell is wrong with ME?

(bwahahahaahahahaha....ROTFLMAO...)

Yes. Yes, I am.

Not to mention that it's a full 8 hours AFTER I had asked "What is LLC?". So yes, yes, I am.

(Trying to understand how you could write a response such as that, without looking at your own lost just immediately above it and see that a) you expanded it and b) then within the same post, you used the acronym. Like, how is that even possible to miss that??? I don't get it. Please someone, explain that to me. You WROTE the damn post. It's your own freakin' post. How can you have missed that? I mean I can understand if someone else wrote it and then you happened to have skimmed over it or something, but it's your own freakin' post. I dunno man. I dunno.)

Course, you're probably reading this and going off on yet another "he's picking on me" rant or some like that. Or lemme guess, you're going to write another one of the "don't tell talk to me like I'm dumb" posts, right?

or maybe...it's just a simple, honest mistake. But man...you make it sooo easy. I don't have to TRY. Like wow...

Difficulty: NOT Asian.

*sigh...*

It's so tough to live by the words of Kipling and Churchill. *sigh...*

[BeepBeep2]

"Today at 09:43 AM" which is AFTER "Yesterday at 11:22 PM" when you KNEW what I was talking about and abbreviated Load-Line Control into the acronym "LLC" yourself.

What the hell is wrong with you?

And if that WASN'T your point, in that LLC only exists in the manual as the expansion of the acronym, you could just have easily pointed me to acronymfinder.com for the same purpose.... At which point, at least I would know ...but not...what it means or what it does (much like how I have expanded CFD and FEA, but probably doesn't tell you much about what it is or what it does).

*edit*

actually....LLC does NOT appear in the manual at all. Do a search for it. (Disproved.)


Now, had I known that it LLC is load line control, and that's how I know to search for say..."load line", THEN it shows up (first instance at p. 44). But that requires knowledge beforehand.

And for computers in general ... LLC could also be an acronym for logical link control.

See: http://en.wikipedia.org/wiki/Logical_link_control

(Course, I'm going to bet that you're going to say something along the lines of "this is just more random that's irrelevant, except that it illustrates how LLC can be referring to other things and the inherent presumption that just cuz YOU know what it's loadline, does NOT necessarily indicate or predicate that other people do. ) --- which I'm sure now you're going to call THAT "nitpicking". You're like a woman (or an environmentalist) - you can never win with them. If it's not clean air, it's clean water. But it's always gotta be SOMETHING that's wrong.

...
I don't know about you, but have you ever noticed that whenever people say that they want someone else to stop doing something, they're often the culprits of the identical or similiar offense? Think about it. Or just sit and observe.

Re-read both threads in case you need a reminder of the manner of your responses.

In fact, from the initial thread - do you want to know what's the FIRST thing you said to me?
...
So, once again who's insulting who now?

*edit* Damn...I am what I hate because I'm taking random puck shots now. (Too easy.)
...
...
And it is PRECISELY for this reason why I NEVER assume people know what I know unless they indicate otherwise. ...

Ahhh...so that's what that is. I dunno. I COULD be wrong, but I think that LLC in that case/sense then has been in workstation/server boards for a really long time then. Probably for the very same reason (at least from the load/unload perspective) although for workstation/server systems, they're probably wanting to squish out fluctations for longetivity reasons. And considering that my Tyan quad Socket 940 Opteron system is still handling the bulk of the simulation work right now still - being a system that's 5 years old, it seems to make sense that they would implement that or something like that much earlier on.

You are obviously the "woman" here.
...

As I recall, that was AFTER you had explained it to me, was it not?

http://www.xtremesystems.org/forums/...=1#post5090219



"Yesterday 10:58 PM"

Furthermore...



From the same post.

Feel free to the correct my logic, but 11:22 PM is > 10:58 PM is it not???

(Christ! You make it so easy!)



....after you...



(bwahahahaahahahaha....ROTFLMAO...)

Yes. Yes, I am.

Not to mention that it's a full 8 hours AFTER I had asked "What is LLC?". So yes, yes, I am.

(Trying to understand how you could write a response such as that, without looking at your own lost just immediately above it and see that a) you expanded it and b) then within the same post, you used the acronym. Like, how is that even possible to miss that??? I don't get it. Please someone, explain that to me. You WROTE the damn post. It's your own freakin' post. How can you have missed that? I mean I can understand if someone else wrote it and then you happened to have skimmed over it or something, but it's your own freakin' post. I dunno man. I dunno.)

Course, you're probably reading this and going off on yet another "he's picking on me" rant or some like that. Or lemme guess, you're going to write another one of the "don't tell talk to me like I'm dumb" posts, right?

or maybe...it's just a simple, honest mistake. But man...you make it sooo easy. I don't have to TRY. Like wow...

Difficulty: NOT Asian.

*sigh...*

It's so tough to live by the words of Kipling and Churchill. *sigh...*

WTF is wrong with you? WTF am I making easy? You look like an idiot trying to correct me, furthermore about something I have more knowledge than you about, every time...

Then you come up with a ****ing racist comment like "Difficulty: NOT Asian"...I am an Asian-American and what you said is extremely offensive.

[BeepBeep2]

Cool. So here's another question then -- why don't CPU/board manufacturers make it like that to begin with then? I mean, I'm guessing that having a lower Vdroop is MUCH better - so why don't they just make 'em that way always?

Intel designed the CPU's to this way, manufacturers have to design the board to be able to meet Intel specs (or something like that).

Im sure someone else can explain this better.

Hmmm...interrresting. I wonder why...I'm pretty sure that there's got to be a reason why they would do something a particular way. At least most of the times anyways...

I already explained this in regards to the transient overshoot upon load when LLC is applied to correct vdroop upon load.
Voltages can spike outside of working specification during the overshoot.

Does everything I tell you just go "in one ear and out the other"?

[alpha754293]

"Today at 09:43 AM" which is AFTER "Yesterday at 11:22 PM" when you KNEW what I was talking about and abbreviated Load-Line Control into the acronym "LLC" yourself.

What the hell is wrong with you?



...

WTF is wrong with you? WTF am I making easy? You look like an idiot trying to correct me, furthermore about something I have more knowledge than you about, every time...

Then you come up with a ****ing racist comment like "Difficulty: NOT Asian"...I am an Asian-American and what you said is extremely offensive.

1) http://www.youtube.com/watch?v=WexWgPqoNF0 One of the top comments was "Difficult: Asian". You'll get it once you watch the reference.

2) When did I start OCing? According to the point-of-sale (POS) receipts, the bulk of the hardware was purchased on 4/17. But I couldn't actually get the system up and running for the lack of a SATA optical drive, which I purchased on 4/18.

3) When did I start this thread?

4) And now, finally when did you mention/expand/explain what LLC is? Yesterday (4/22). That's FOUR days later AFTER I had started OCing.

5) If your point of bringing up "read the manual" was an attempt at "RTFM" - as indicated/stated, I DID RTFM. And without what LLC expands to BEFORE yesterday (remember that I would have started OCing FOUR days PRIOR - so reading the manual FOUR DAYS before you even explained what LLC is, wouldn't have help at all).

6) When did I hit the 4.5 GHz mark? Sat Apr 21 sometime between 12:33 PM and 2:15 PM.

So....to recapitutate the sequence of events (since apparently you're incapable of logical thought)

4/17 - bought hardware
4/18 - bought SATA DVD drive
4/19 - thread start
4/21 - hit 4.5 GHz
4/22 - you finally told me that LLC is load line calibration (because I missed it from someone else the first time)
4/23 - today.

So how is knowing what LLC is a day AFTER I had hit the 4.5 GHz OC target of ANY use to me, ex post facto???

And then to imply that I didn't read the manual when I very well did. Nice try though. *Stand O*

You make yourself look like an idiot.

Why on EARTH would you say "You abbreviated it yourself." when all evidence proves the contrary? Are you so incapable of admitting that you are wrong? (I mean, not that I really care...but I'm curious.)

See, here's another one of those "incomplete" things. If you're going to say something like that, check yourself to make sure that that's actually a true statement before saying it. It'll prevent the embarassment of my having to illustrate and point out and cite and demonstrate your idiocy by simply proving that you're wrong.

And the fact that I am assuming that you didn't bother to check, makes it THAT much easier for me to do JUST that.

You also DO realize that people who SAY they have more knowledge about something often don't actually have to tell people that right??? (Because if they did have more knowledge about a particular subject matter, what they have to say about would become inherently apparent.)

For example, I had just asked about why wouldn't they make boards/chips such that the there's minimal or negligble Vdroop, and I've yet to hear you chime in on that. Instead, you're still b!tch1ng.

Now, for someone that says they know more than I do (and I don't doubt that at all...not in the slightest - you've been doing this a helluva lot longer than I have), I would have assumed that you would be far more interested in disseminated the technical knowledge/information because that's what you thought to be more important. Apparently, I'm wrong. Again. ( spit happens.)

You also DO realize that by saying something like "I know more than you do" means that the expectations are going to go WAYYY up now and that you're just opening yourself up to the whole kind of "I thought you said you knew more than I do" type remarks. You DO realize this, yes???

To me, I don't really care if know more than I do or not. Clearly, if I'm declaring at the very beginning and throughout that I'm NOT an ECE nor an OCer, that's pretty much a redundant, moot, and obvious statement. So, thank you for that, Captain Obvious. (Aspiring to be a patent lawyer much? cf. KSR v. Teleflex). (and yes, there's a reason why they're called parenthicals, which I'm sure that you'll skim over and write-off as irrelevant and miscellany BS.)

(In case you can't tell, based on your remarks, I don't particular think very highly of you.)

Boy...you're going to make a GREAT prof someday. That way, you can begin EVERY semester with your whole "I know more than you do" remark as you babysit a lecture hall full of freshmen.

haha...like...why do you even say stuff like that? I mean...I don't even have to make it up! This is GREAT feeder material!!! LOL...

(hence why I said - too easy.)

And don't...bother with a comeback. I'm not interested. It'll in all likelihood, just end up being MORE feeder material.

And I really don't care if you think I'm smart or dumb. Because like an election, and to quote George Carlin: "...I'm sure as soon as the election's over, your country will improve immediately." (i.e. Tomorrow's just another day, smart or dumb. I really don't care which or what your opinion of me are in that respect.) Course, you're probably sitting there, skimming this going "WTF does THAT have to do with anything?"

You've clearly never been on a debate team or in a debate in elementary school. Point. Counterpoint. It's that simple/easy. Or in your vernacular, "nitpicking". (boy...I'm racking up the mileage on that one. haha.)

P.S. I'm Asian too. What's your point?

Watch the video. Then read the top comment: "Difficult: Asian". It'll become self explanatory then. And then it'll also become self evident why I referred to you as "Difficulty: NOT Asian"

(I'm pretty sure that if I were to retell this story, some people aren't going to believe me - as if I'm actually making your responses up.) lol...oh well.

Anyways, back on topic - running a different kind of FEA now. I forget if the thing run even converges and/or how long it takes for it to converge. Implicit FEA, with non-linear materials.

Using 10 GB of RAM out of 16 GB. Total I/O traffic so far - ~809.4 GB. (Give or take a few GB). Thank GOD I'm running it on the SSDs.

[alpha754293]

I already explained this in regards to the transient overshoot upon load when LLC is applied to correct vdroop upon load.
Voltages can spike outside of working specification during the overshoot.

Does everything I tell you just go "in one ear and out the other"?

That says NOTHING to why Vdroop isn't zero or near zero.

You just explained how LLC works.

Obviously the Vcore@load can be very VERY close to Vcore@idle.

So, why would you have to SET the LLC to "ultra high"? Why wouldn't you just MAKE it that way?

The analogy to that from my world is that we can't make everything fit together perfectly because of manufacturing tolerances. Ideally, you should have zero gap where they're supposed to be zero gap, but we might only get that when one is at the their max tolerance and the other is at their min tolerance. (or something like that). Which means that by design, there will be a nominal gap in there.

So, is that why the LLC isn't at "ultra high" by default even though it would (presumably) give the better, more ideal condition where the Vcore@load is almost equal to Vcore@idle?

(Clearly if I'm asking it again, it's indicative that maybe I haven't fully grasped the concept yet.)

And with the transient overshoot, it's a voltage drop at load, so what's the opposite of a voltage spike? It'd be like a voltage trough right? Load goes up, voltage goes down. Transient overshoot would be downwards would it not (wouldn't that technically be an undershoot then?)? (Because you start off at a higher voltage and then you go DOWN to a lower voltage...)

I dunno.

No wonder why I had to take circuits for dummies twice.

And if the voltage overshoot is bad, then why would OCers want to use LLC such that the voltage difference between idle and at load would be minimal?

See, that's where I'm confused.

You want min(Vcore@load-Vcore@idle), and you can get that with the LLC. But then you're saying that "voltages can spike outside of working specifications during the overshoot". So on one hand, it is very good to have minimal differences between that. But on the other, spikes can exceed the working specifications during the overshoot. (And that would be bad why exactly? Would it fry the chip or something? Or some how otherwise damage it?) I still don't get why OCers would want to minimize it if the spikes can be such a bad thing. Is that just a risk that OCers are willing to have to take in order to get the two Vcores to be MUCH closer together?

P.S. The way that you answer questions reminds me of nearly EVERY SINGLE PERSON that I've ever met that's a programmer. Ask them why something works the way it does and they're like..."I dunno. I just copied and pasted and tweaked it till it works. Don't ask me HOW it works. I just know that it does.

P.S. #2 You also remind me of the people that modify their cars. They lower it and they're like..."yea...I got this shock, and that spring, and I dropped it 2 inches." But when you ask them a basic vehicle dynamics question like "where's your new roll center and your instant center?" They have noooo friggin' clue. And actually....they often respond a LOT like you do. With great discontempt. (For an otherwise seemingly simple question that really only necessitates a simple answer.) It's no freakin' wonder why I never went into ECE. (Yes, I am assuming that you're an ECE major).

[BeepBeep2]

Can you read?

What I said was very clear.

[alpha754293]

Can you read?

haha and then you wonder why I answer you the way I do. lollllll and then cry about it. lolllllll

And then you wonder why I end up disproving you. Repeatedly. lolllll you are a effing riot. lolllllll And then you /cry about it some more. "mommy, mommy he's picking on me." lolllllll wow.

And then you want to say that you're all knowing and helpful/useful? How is THAT ANY BIT helpful at all???? Seriously??? And you want to what? And you want me to treat you as a foremost authority on overclocking? WHAT???? Are you kidding???? Wayyy to demonstrate your whole "I know more than you do" there bud. lolllllll

It funny how you seem to think (or at least that's the impression that I get from you) that "ohhh you're such a valuable asset and you've helped soooo much" when the only contribution you've made thus far is to explain to me what load line calibration is to me. THAT's IT. So get over your ing self.

[BeepBeep2]

You don't even disprove me...you just about .
I could have neglected to help you at all, where would you be?

I have NEVER claimed to be all knowing. I said I know more about the subject at hand (overclocking, d******) than you. You constantly attempt to refute my claims with silly accusations like "Your answer was incomplete"..."If only I had known"..."If you would just have explained" about things you didn't even know about until I told you, in which case you didn't comprehend it anyway.

Why do I need to explain explanations? Look into it yourself or ask questions instead of insulting me.