EVGA

Helpful ReplySpecific questions regarding EVGA's LLC-settings

Author
Philkingz
New Member
  • Total Posts : 4
  • Reward points : 0
  • Joined: 2019/03/11 06:00:05
  • Status: offline
  • Ribbons : 0
2019/04/22 09:22:50 (permalink)
Hi there,
in this thread i want to ask some things about EVGA's specific LLC-settings (edit: for the EVGA z390 Dark).
 
While i wait on the rest of my hardware I've spent weeks on grinding through articles and videos to understand how Vdroop and LoadLineCalibration actually work. I think I reached a point at which i start to understand a bid more than the typical phrases, which also led to frustration as i realized how distorting and misleading a lot of videos and articles from big Youtubers and official Wiki-site are; at least for a layman like me. I read some posts from Delirious and those are clearly in the category "helpful", since he seems to have run into similar problems with understanding the settings as i do. So i quess this forum is the right place to ask my questions.
 
 
-----------------------------------------------------------------------------------------------------------------------------------------------------
 
1) Vdroop settings are made in percentages. As i understand there is a "natural" vdroop already implemented in the power specifications of the systems in order to save power, stay within TDP specifications and have better load transients (even though the last reason is often overly exaggerated as its main purpose.)
So i just want to make sure when EVGA states "-75% vdroop" that it really refers to the inherent existing vdroop as basis for a change of 75%. Is that correct?
In other words: With no LLC ("0% to vdroop"), there is still a vdroop, right? So the EVGA settings are not refering to some absolute values that simply change the voltages under load, but to a percentage change of the existing drooping, which in turn changes voltages in correspondance to current load. Is that correct?
 
 
2) The next one is simple:
"-X% vdroop" would decrease the vdroop effectively letting the voltage drop (that corresponds to the current load) be smaller. Hence it allows to aim for higher voltages under load. "+X% vdroop" on the other hand should make the vdroop stronger. Hence letting the voltages under load drop even lower. Im quite sure after reading some posts on the forums that this is correct, so a simple yes is enough.
 
 
 3)  I also read on the forums that the LLC-settings even apply voltages to idle state. At that point im not sure if i have understood LLC and Vdroop completely or if i misunderstood the posts. So my question splits into two different ones.
 
a) If you change the load-lines, shouldnt the voltage changes at idle be a lot smaller than the changes to voltages at load?
I mean calibrating the load-line would not be a "parallel shifting" of the line, but a change of the "steepnes" of the curve, correct?
Professional MS-Paint-Picture:    ht tps://imgur.com/a/3FERtbQ      
(since i cant post the link directly, you need to manually remove the space character)
 
b) If the upper one is correct, then the changes to voltages at idle should actually not be very high, despite some posts i've read.
So is there something different in EVGA's LLC settings than in others manufacturer's settings  or  do i still have not completely understood how LLC and Vdroop works?
 
-------------------------------------------------------------------------------------------------------------------------------------------------
 
 
Many thanks in advance for any helpful responses!
post edited by Philkingz - 2019/04/22 22:40:45
#1
bcavnaugh
The Crunchinator
  • Total Posts : 35026
  • Reward points : 0
  • Joined: 2012/09/18 17:31:18
  • Location: USA Affiliate E5L3CTGE12 Associate 9E88QK5L7811G3H
  • Status: offline
  • Ribbons : 226
Re: Specific questions regarding EVGA's LLC-settings 2019/04/22 09:26:01 (permalink)
Welcome to the Forum
On what Motherboard?

  Associate Code: 9E88QK5L7811G3H
"Earn EVGA Bucks For Folding with EVGA!" it is not "Fold for EVGA Bucks with EVGA!" “Do it for the sake of it"



#2
Philkingz
New Member
  • Total Posts : 4
  • Reward points : 0
  • Joined: 2019/03/11 06:00:05
  • Status: offline
  • Ribbons : 0
Re: Specific questions regarding EVGA's LLC-settings 2019/04/22 09:38:34 (permalink)
Thank you for welcoming me,
sorry that i did not clarify on this: It's about the EVGA z390 Dark, which is resting beside me while the rest of the hardware is not, still waiting for it. So i cannot take a look into the UEFI myself. I can only watch vidoes and read through articles/forums to prepare and understand the basic stuff i have to deal with.
 
#3
TiN_EE
EVGA Forum Moderator
  • Total Posts : 247
  • Reward points : 0
  • Joined: 2010/01/22 21:30:49
  • Location: xDevs.com
  • Status: offline
  • Ribbons : 13
Re: Specific questions regarding EVGA's LLC-settings 2019/04/22 20:53:33 (permalink) ☄ Helpfulby Philkingz 2019/04/22 22:20:16
"Default" condition is (LL) Loadline vdroop that we preset as stock shipping configuration, selected based on our power, CPU stability and thermal testing. Each vendor and each board design have different implementation of what is considered "default" loadline. That is a reason why comparing loadline vdroop between different boards/models/manufacturers is pointless task, unless very careful and elaborate measurements are made to determine actual data. Main goal of loadline vdroop implementation is to reduce stress on CPU(GPU) under load.

Because LL is not a standard industry thing, there is no "specification A" or "definition B" that everybody trying to follow. That is why everybody, both reviewers, users and manufactures use their own terminology in this, which does not help to clarify things.

1) Yes, 0% setting at EVGA BIOS still have some droop, it's just kinda in between the other settings. Absolute values like volts or milliohms wouldn't help much on matter, because actual loadline droop can vary a lot depending on installed CPU, used voltage levels and workload. Light benchmark will have small voltage change, while heavy AVX workload will be way more severe. That is why loadline called so, it depends on load. So your understanding is right.

2) Yes. However this also affects power and temperatures a lot, resulting clear difference in overclocking and thermal results.

3) VRM controller cannot predict future, so in reality setting more stiff loadline (less slope, to get smaller voltage droop, negative %'s in EVGA BIOS) would also change idle state voltage. By such change VRM controller "anticipates" upcoming load, to prevent voltage to swing outside of the margin when that actually finally happens.

a) That actually a lot depends not on loadline setting itself, but on implementation of hardware design on particluar motherboard.

Let me do some measurements and provide some examples here.
I apologize for all Z390 DARK owners, I should have addressed this important aspect earlier.
#4
Philkingz
New Member
  • Total Posts : 4
  • Reward points : 0
  • Joined: 2019/03/11 06:00:05
  • Status: offline
  • Ribbons : 0
Re: Specific questions regarding EVGA's LLC-settings 2019/04/22 22:20:01 (permalink)
Hello TiN,
thank you very much for taking my concerns seriously and responding to them in such a direct and concise manner.
I'm looking forward to see your results.
 
 
------------------------------------------------------   Offtopic    --------------------------------------------------------------
ps. If there is Ilya Tsemenko behind your alias, let me also thank you for providing such insightful information about powermanagement  and -specifications of GPUs at GamersNexus#107. The entire industry seems so bloated with advertisement through fancy pictures and blingbling that spreading information about basic, yet essential, stuff like powermanagement, wear and reliability of hardware has completely disappeared into the background. I feel even better now spending that 550€ on a quality board, 'cause i know there are people behind the product that try to work against this trend. Greetings from the other side of the earth.
#5
TiN_EE
EVGA Forum Moderator
  • Total Posts : 247
  • Reward points : 0
  • Joined: 2010/01/22 21:30:49
  • Location: xDevs.com
  • Status: offline
  • Ribbons : 13
Re: Specific questions regarding EVGA's LLC-settings 2019/04/23 00:11:35 (permalink)
Yes, I'm that TiN. Nickname is just my full name initials, there is no relation to metal name :D.
I think complex matter of modern chip operation and NDA regulations is big part of the reason why little clear information is available in public.
 
Even simple things like loadline and vdroop effects or power management are complex result of many variables, temperature, voltages, currents, physical design (like PCB layers, used components, etc.) and VRM circuit design. There is no universal rule like "Use X layers PCB, Y phase VRM and Z type capacitors" to get best result, each case need optimal solutions for desired performance/cost/formfactor. Perhaps you can also find helpful my recent update about VGA's power limiting and some detail about loadline scheme, published in RTX 2080 Ti Kingpin guide. While it's in GPU flavor, idea there apply to motherboard design as well.
 
#6
Philkingz
New Member
  • Total Posts : 4
  • Reward points : 0
  • Joined: 2019/03/11 06:00:05
  • Status: offline
  • Ribbons : 0
Re: Specific questions regarding EVGA's LLC-settings 2019/04/23 01:42:15 (permalink)
I understand what you say with there is no universal recipe for every dish, but there should be an inherent, basic principle to make out. I mean that is the reason why you gave me the link to the RTX-Guide, as i will be able to understand the principle and transfer it to other applications. Yet, in terms of LLC and Vdroop there is really much information out there, which is detrimental to a beginners understanding of the principle, at least to mine.
 
 
As an example take look at the article of the official wikichip.org ( ht   tps://en.wikichip.org/wiki/load-line_calibration ). I dont want to elaborate on this too much, as it is very hard to explain what i find misleading without becoming very complicated myself. I also fear i would get into serious trouble with some people, if I'm not fully correct with my objection. It has to do with my impression that they seem to not exactly differentiate between inherent vdroop and the natural ripple and noise due to changes to the duty circle of the VRM, which tries to compensate for a voltage drop. It seems they either somewhat combine it in their initial pictures or they draw some unrealistic pictures without declaring the assumptions its based on. Im not sure if there is an inherent vdroop implemented in the second picture or if they really think there would be a permanent voltage drop caused by increased current load, that has neither ripple or noise and gets not compensated by the VRM. Those factors (ripple and noise) are merely depicted as a result of the reaction of the VRM which changes its duty cycle. This is partly true as the ripple is the result of the switching frequency of the VRM, but that is exactly the reason why there is always a ripple and even some noise, whose strength is dependant on the current changes as well as the factors reducing the strain of the current load like capacitors. It is not that there would be some natural voltage drop corresponding to the current load that gets not compensated  or  that only compensating from the level of a lower voltage results in excessive noise. In this regard it's important to differentiate between intentional voltage drops (vdroops) and short dips to make clear how changes in the duty cycle affect the voltage level and where noise is situated in contrast to normalized ripple. Perhaps im just not getting their didactic approach and they actually try to explain that in those pictures, even though then there is something wrong with their subsequent explanations. Maybe you can grasp it better than me.
 
 
Anyway, Im sure the RTX guide will help me, but i need to sleep before i can take a look. Thanks, again.
post edited by Philkingz - 2019/04/23 10:28:42
#7
TiN_EE
EVGA Forum Moderator
  • Total Posts : 247
  • Reward points : 0
  • Joined: 2010/01/22 21:30:49
  • Location: xDevs.com
  • Status: offline
  • Ribbons : 13
Re: Specific questions regarding EVGA's LLC-settings 2019/04/24 02:54:40 (permalink) ☄ Helpfulby Philkingz 2019/04/24 10:42:00
In loadline and overall power delivery understanding it is very critical to establish the bases and define on where we look:
* Voltage delivered to the CPU cores at the die? < this we cannot directly measure, only estimate from software readouts.
* Voltage delivered to the CPU package (pins)? < this we can measure, but few are going thru effort.
* Voltage delivered out of the VRM (inductor output)? < this is what usually reported by reviewers.
Depends on where you tap the reading, picture on droop will be very different.
Also duty cycle is not the only thing that affects ripple and transient performance, there is also phase shedding on low loads (disable phases to improve efficiency).
#8
Jump to:
  • Back to Mobile