How load line calibration actually works

[@NovemberOrWhatever]

Every AHOC video: I know it sounds complicated, but it's actually even more complicated.

[@bananya6020]

Buildzoid drew an actual diagram for this video. this shows he is truly serious about fixing his mistakes

[@PwadigytheOddity]

"I was wrong," "and here's why I was wrong," "let me make drawings correcting what I was wrong about so you know the right information" - absolute and utter Chad: Builzoid Something sorely lacking in the tech space is being unable to directly state "I was wrong"

[@ActuallyHardcoreOverclocking]

First So this just got to me from a more reliable source than I am. The SVI2 TFN on Ryzen literally just reports what the VRM measures. You should probably watch this: kzbin.info/www/bejne/rpzIg2SQa5eqn5o

[@theophilusthistler5885]

Your comment has potential.

[@The_Man_In_Red]

Wait that's cheating

[@ActuallyHardcoreOverclocking]

@@The_Man_In_Red I post my first comments after going through the YT page for my channel. So it's not impossible to beat me. Just really unlikely since I know that the video is public first.

[@dugg117]

hi

[@spike229]

5th!

[@Vladek16]

I think you should delete your old video on LLC if they are wrong. Especially now that you've made this video on the subject :D

[@andrewryder3075]

I think we're talking about two different things here. My understanding is that Active Voltage Positioning, which is what you were describing in your previous video ("Transient Response in Motherboards and GPUs part 2: load line / LLC /AVP") takes place in the PWM controller, while LoadLine Calibration is actually a calculated BIOS response to combat droop between targeted and produced voltage - at least that's how it was described to me by a Gigabyte engineer several years ago. (If I was misinformed, I apologize for the earlier comment).

[@ПётрБ-с2ц]

Basically, BZ says you have good enough overshoots and undershoots at "high" setting. So if a motherboard wanted to correct the load voltage without introducing overshoot/undershoot it should use AVP. The big problem is, it should do that with at most as large latency as VRM does react to CPU voltage request. Does the CPU request voltage directly from VRM or does motherboard relay it?

[@kian8382]

BZ‘s confession about his past mistakes is the only thing that melts a keyboard warrior's heart, don't worry, we believe in 2nd chances.

[@brovid-19]

Wish my ex was as forgiving

[@patrickfouhy9102]

You are awesome. You are the only person on youtube I've found that actually updates information. Keep it up!

[@ExPsy]

Being wrong every now and then is part of the process of growing as an engineer. The whole idea is to work around your wrongs as they accumulate. If that wasn't the case every thing we made would be perfect and there would be no need to make it better.

[@warrenpuckett4203]

If it was made better there would be no need to replace it. I have often said to many who don't want to understand this. Why do Paccar corp vehicles most often have inline 6 power plants and your purse hauling, grocery getter most often has a a V type engine or a I four with turbo? I have a Win 2K Pentium 3 on the floor now that someone wants me to fix.

[@stoneymahoney9106]

This video should be mandatory watching for EVERYONE from now on before they're allowed to overclock ANYTHING EVER AGAIN.

[@soldierbyname]

Basically figured out what you are explaining just recently while overclocking my 3800x, just no where to your detailed explanation. Love the videos. Thank you very much. Now I have to go back into bios and mess around again haha. Posted on many forums to figure out how to actually measure vcore and well you are the only one who answered it with a proper technical explanation . Picked my x570 solely after watching your round ups. Keep up the good work.

[@OlettaLiano]

What did you get your 3800X overclocked to? I have mine overclocked to 4.5 GHhz @ 1.375 volts on all cores.

[@soldierbyname]

@@OlettaLiano 4.4ghz @1.256vcore, llc3 or med/high. At the moment limited by my 2070 so down the road 4.5ghz ++ will happen

[@OlettaLiano]

@@soldierbyname I played with LLC but I get better benchmark numbers with it on Auto.

[@soldierbyname]

@@OlettaLiano my experience is the opposite so far. In 3d mark and r20 my scores are bit higher. Im running tuf x570. Mobo might makes a difference.

[@OlettaLiano]

@@soldierbyname I'm running a 5700XT and both my CPU and GPU are water cooled with an open loop with 2 360mm radiators and 12 120mm fans in push/pull. My case also has 2 140mm fans on the top, and two more on the rear panel.

[@Xerpadon]

FINALLY! i've been people telling this for years! now i can finally point them to this video!!!

[@usrname1y]

You are the OC Sensei I've been searching for my entire life 😁

[@TheCeph]

Your videos have single handedly helped me understand, diagnose, optimize and straight up fix issues with my pc. Youre amazing man. 🤖

[@tcclaviger]

Probably the most useful OCing video I've seen in 20 years...

[@cliftonsr]

When it comes off load is where it spikes a bit and then comes down so your trying to compensate with LLC for excessive droop voltage and it will overshoot a bit if you look at your graphs while loading certain workload calculation. So the higher number settings you enter is good to compensate for droop but watch out with v-core plus this slight voltage spike overtime could cause some degradation for sure. But LLC can be your overclocking friend for stability. Been in the computer builds since the 90's. Great to see young guys like you in the field that's why I subscribed and would recommend others to do so as well and keep up the good work BUILDZOID!

[@dusliangames]

I love these kinds of videos.

[@Alaakso]

This explains why the different voltage readings (cpu vid, sio vcore and vrm vrout) are so far apart. Now I know that I only look at my IR35201 readings. Gigabyte z390 i aorus gladly let's me read these through hwinfo. For example now @idle vrout is 1.254 and vcore is 1.26. Thanks BZ!

[@lagynas]

Buildzoid: don't blame him. Dirty casuals: attack!!!

[@lukasem7024]

Fantsastic video mate ! Many many thanks

[@fablearchitect7645]

1:25 so the video "Explaining Load-Line Calibration (LLC) & Not Killing Your CPU" on gamer nexus from 2015 is wrong and is still up spreading disinformation?

[@cmdrsoyo7154]

nothing starts my wednesday better than a new Buildzoid video

[@EvilAngelix]

Someone call Steve from GN, there have been shifts in the data!

[@spamlucal]

So that's why SVI2 TFN sensor reports almost the same as Core VID on HWInfo in high end Asus motherboards.... This may be a silly question, but... Why does HWInfo report different voltages for each core? As long as I understand, there is just one voltage going to the CPU from the main VRM (this is, ignoring SoC voltage)

[@ActuallyHardcoreOverclocking]

every core can request it's own vid but the VRM only produces one voltage so which ever VID is highest is the one that gets applied to the whole CPU.

[@Thex-W.I.T.C.H.-xMaster]

Love your videos you rumble sometimes but I understand what your getting at in most cases 😅.

[@Hugh.Morjowie]

So how loadline actually works and what actually Vdroop is? I didnt get it from video.What LL trim actually does? What connection it has to actual voltage, how it affects that??

[@asaadb1]

LLC: How much throttle you keep while your foot is on the clutch.

[@warrenpuckett4203]

Yes but using your nose to calibrate is gets expensive.

[@magnaxilorius]

I just set my LLC setting to what gave me the closest to my VCORE set in BIOS and stress tested from there.

[@Brometheus420]

So you helped me understand how it works, but then said throw that out the window if you have an asus mobo lol. I have a Asus Z590 gaming wifi-e and an 11700k but now I don't know if I should use high LLC or extreme. If 1.3v is good for daily use then 1.4v with high llc should be 1.3v under heavy load? Or would it be better to just set it to 1.3v and extreme? on OCCT I can't figure out which are the VRM controller or super IO, but Core 0-7 VID shows max spikes of 1.45-1.465V and lows of 0.66V so im guessing that is the super IO or the stock settings just let it get zapped?

[@ralphbarnes9335]

Good and informative video. Thanks

[@berberger4814]

its actually so easy when you know whats happening

[@JorvsUchiha]

when will make a how to under volt and LLC at the same time? seen some video done on 75950x3D or newer amd cpu i dont know if it safe?

[@1300l]

Nive, very well detailed

[@mythbuster4315]

Thanks bildzoyd

[@tresnugget]

I'm curious if using offset vs manual voltage input effects the amount of LLC you need. I have a Maximus XI Apex (started with the code and my 9900k went from requiring reported 1.28v for 5 GHz to 1.27v for 5.1 under load). Everyone said to use LLC 7 for manual but LLC 5 seems to work great for offset. It looks like there's a tiny bit of droop

[@ActuallyHardcoreOverclocking]

LLC7 is way too much LLC on ASUS boards

[@GurkenkoenigZ]

@@ActuallyHardcoreOverclocking yep on my Asus z170 ws board lvl 3-5 is pretty Spot on Depending on the powerdraw

[@alexandermyrthue1987]

Sooooo if I can't ask the VRM controller to spit out a voltage measurement because of analog vs binary then how can I command it to give me x amount of voltage?? 🧐 At least this must mean that when I ask for lets say 1.45V the controler just have to guess what I'm trying to say, right?

[@jake20479]

...my mind is blown. honestly. everything I "knew" about LLC is out the window.

[@spamlucal]

QUESTION: If you have no idea how your motherboard is reading Vcore, can you still rely on it? is it accurate in its inn-accurateness? example: If a sensor reports 1.17V (but actual Vcore is 1.08V), and you adjust LLC and offset voltage settings, and it still reports 1.17V, is the underlying actual Vcore the same as before (1.08V in this case)?

[@DrakkarCalethiel]

Never thought that the DIE-sense circuit is so damn senitive!

[@ActuallyHardcoreOverclocking]

well the controller needs to be able to see the transients caused by the CPU. So while the VRM won't tell you the mins and maxs of the voltage during transients it does react to them

[@Fix_It_Again_Tony]

The connection inside the CPU is probably just a connection directly to the Vcore voltage. Not very sensitive. You could sense the Vcore voltage right at the VRM output but then you have the impedance (capacitance, inductance, resistance) of the Vcore traces going to the CPU which would cause error in your measurement (look up "four terminal measurement"). You want to sense the voltage directly at the processor, not at the VRM output. The processor doesn't care about the voltage at the VRM output, it only cares about the voltage on the processor. This is what the VRM needs to sense to be able to control Vcore accurately (accurate from the perspective of the processor). The VRM feedback (commonly referred to as Vsense or Vsen) is a high impedance node, that is it doesn't load the signal very much. It means it does not draw much current so if you connect this directly to the Vcore at the processor you don't have the same error in the measurement. However, this presents it's own problem. Because it is high impedance it is susceptible to noise. This is what he is saying at 7:32 when he talks about running the voltage sense signal next to a fan header. Because it is high impedance it also picks up noise easily. You want to pay attention to how you route this signal to keep it away from digital noise. This is a problem with all DC/DC converters using external feedback. Also the scope probe itself is not perfect and loads the signal with it's own impedance. This may affect the gain and phase of the signal back to the VRM. This always reminds me of the "measurement problem" in quantum physics. These two phenomenon aren't caused by the same mechanism, but the effect is similar: the act of simply measuring something can change it's state. It's not uncommon to be troubleshooting electronics and when you hook an oscilloscope to a signal the problem goes away. Disconnect the scope and the problem comes back. The scope is affecting the signal you are measuring. This is usually caused by the scope probe impedance and not external noise. There are special high impedance probes, but they generally cost a couple thousand USD.

[@KaTooMaz]

Well...Ummm...Yeah...Soo...Yeah... Thanks for the vid and info!!

[@rubenbuckens399]

Hey buildzoid, small question. When you're doing a video like this, where do you find your sources?(Like for example, where can I find the document that you're basing this on?)

[@jesperjrgensen2184]

So, how can I determine wether or not my transient response is better or worse for a given LLC - Any rule of thumb that can be applied? ASUS Z390. LLC 5, VCORE 1.360v - VCORE @ LOAD 1.323v. LLC 6, VCORE 1.360v - VCORE @ LOAD 1.367v. Obviously LLC 6 is causing a vdroop, but is not showen in windows because the Super I/O readings are wrong. But how can I tell whichever have the best transient response without an oscilloscope ? :D

[@rnrkdqls86]

damn this video gives me a whole new perspective in overclocking. i'm using a phantom gaming itx for my daily rig with a 99k that is intelburn stable at 1.3V idle 1.31V at full load @ level 1 LLC settings. the level 2 LLC droops the load voltage to 1.26 so i would have to add extra volts in the bios to pass the intelburn at load and it would yield my idle voltage to be around 1.35V with load at 1.31V. So u're suggesting the voltage readings at hwinfo is actually the super IO readings and it would be best to use level 2 even if hwinfo and cpu-z voltage readings show a greater discrepancy between idle-load voltages? (coz the readings are all bullshit??)

[@stevenwest1494]

Skip in the video 20 minutes for the direct LLC breakdown and info.

[@djc5166]

So the reason all this exists is that VRMs built on inductors cannot physically regulate voltage fast enough when the load on the CPU is changing? So LLC adds some small resistance offset as a function of current draw?

[@suunto61]

Dude, reach out to the processor guys and get their spec. They want that output impedance from DC to light.

[@ActuallyHardcoreOverclocking]

The LL settings for intel are public.

[@MikrySoft]

I always preferred "DC to daylight". Alliterations FTW

[@timeDrapery]

@@ActuallyHardcoreOverclocking Okay, I believe you on Intel LLC settings being "public" however after a good amount of searching I've yet to come up with documentation regarding their settings OR the implementation/settings for any other motherboard manufacturers. I believe this is due to them being "encrypted" using "jargon" (AKA, "engineer talk") which precludes my discovery of them without the help of an adult... 😂😂😂😂😂 TLDR; Where do I find (or, if you prefer, how do I effectively search for) the documentation you're referencing that provided the information presented in this video? Where do I find the documentation describing load line calibration implementation/settings for the major motherboard manufacturer's (such as gigabyte, ASUS, etc.) products? Thanks for all you do, your videos are some of the best presentations of personal computing information that I've seen and I appreciate all your hard work! Take care!

[@bl4ckth0rn39]

ASROCK has a F-STREAM utility. In the system tab, all voltages are displayed in real time . CPU input V., Vcore, DRAM V. ... . I have no idea where these voltages are read from. I hope that directly from the voltage regulator. (Z97 KILLER)

[@younes6933]

Now I'm even more confused. I read Vcore with a digital multimeter off the capacitor bank on the back og the motherboard. There I get a HIGHER voltage reading than Vcore set in the bios under load with "turbo" LLC, while I get a LOWER reading when I set "high" LLC. Now I don't know what number to trust or what LLC I should choose.

[@sidlabdi]

Hi, Thank you for this tutorial. vVery Useful. I have recently built a computer based on the Asus Maximus 13 Hero + intel i9- 11900K. In the Bios menu, I have a setting named "Voltage Monitor" with two options, i.e. "Die Sense" or "Socket Sense" under Extreme Tweaker\DIGI+ VRM. Which one should give the most accurate Vcore reading in your opinion ? Best regards sl910

[@gsuberland]

So the original comment was almost entirely correct, in terms of correctly explaining that LLC compensates for voltage droop due to resistive losses, but was overplaying the overshoot issue because people incorrectly measure the voltage from SIO? And just so we have concrete numbers, I pulled up an 8th gen CPU datasheet I had laying around and it says maximum impedance and capacitance on the Vcc_SENSE / Vss_SENSE lines is 1MR and 1.5pF respectively, mentioning that the ground probe should not be longer than 5mm, and must be shielded. It also says to expect >10MHz on that line. I suspect the parasitic capacitance from a multimeter probe would be sufficient to exceed the 1.5pF limit and act like a low pass, even before you consider coupled noise.

[@GurkenkoenigZ]

Only cool peeps have datasheets laying around. I dont :(

[@gsuberland]

@@GurkenkoenigZ Google "Technical Resources: Intel Core Processors " if you want to download them yourself.

[@ActuallyHardcoreOverclocking]

except load line HAS NOTHING TO DO WITH POWER PLANE RESISTANCE the controller already compensates for that by measuring at the die and not at the output of the VRM inductors.

[@gsuberland]

@@ActuallyHardcoreOverclocking right, so it's drop over substrate?

[@mattfox8212]

MSI gaming pro carbon/ ryzen 2700 @ 4.1GHz. 1.375v bios set, LLC4, SV12 TFN idle - 1.369v, load - 1.316v. Time & Gflops improved from LLC3 (1.344v load) in IBT & tests stable @ LLC4 so far but not R20. Should my load voltage be closer to the set 1.375v? LLC2 drops to 1.369v & LLC1 overshoots to almost 1.39v. I appreciate your content and the time required to create it. Thank you. The concept makes sense but not sure how to correctly troubleshoot how the board is implementing it.

[@gigigigiotto1673]

i consider voltage spikes as overshoot

[@hubertcumberdale8175]

But you say when you go extreme llc, the voltage regulation gets worse. Ok so what does that mean? It kind of sounds like you could get overshoot, and undershoot, by using the extreme setting. So in the end its the same because if they are protecting against vdroop, im willing to bet that the voltage is higher more than lower. So despite all this new understanding the end result is the same as the incorrect understanding.

[@Multimeter1]

28:18 so what about say an EVGA 1080TI FTW3? Is it accurate in gpuz read out?

[@williamwolfe9335]

For someone who just wants to overclock, how do I use any of this information to know which llc to use with my system? I have a z690 msi meg ace and an i7 12700k with 32gb of ddr5. I know i can increase or decrease voltage using load line calibration, lite load settings, or a voltage offset, but i don't know how to decide which to change and how to optimize all three.

[@LecoSilvaPoesia]

Thanks a lot for this video! I was really confused with the LCC behavior before, for I was considering your Gamer Nexus' entry from 2 years ago. I've been using VR VOUT to monitor voltage for a while, now, but I'm still a bit confused with WHAT I should expect as a sustainable, healthy VDROOP. Using both medium and high LLC with a Z390 Aorus Ultra mobo I get that 0.1 mV VDROOP and using turbo get 0.05 mV. Should I aim for a higher VDROOP then? For everyday oc'ing? Using AUTO I was getting overshoot (unecessary 1.4+ for 5.0-5.1 GHz i9-9900KS) but with MEDIUM-TURBO and manual VCORE settings that's not happening on heavy CPU gaming (emulator RPCS3, which I've found to be a GREAT strees testing tool lol) or Geekbench. I don't have the guts or knowledge to use Prime95 tho Thanks in advance for reading this and answering, if you may.

[@arkama67]

Buildzoid it's okay to make cut in the video and also a bit of editing in order to shorten it an be easier to watch. I really think you should try a bit.

[@simpleman9449]

So it’s bad to adjust the ac/dc LL? We should only be adjusting the LL CAL “trim” on modern boards?

[@mindinfinity3714]

If you set a negative PBO like -35 for a CPU like 7800x3D, is it safe changing the LLC setting from default?. I ask because, some people say to stabilize a large negative PBO offset it is necessary to manipulate the LLC setting.

[@n0rie9a]

dude, where were you back in the P45 days? this type of info was like top CIA secrets (also mosfets, VRM ctrlrs etc)

[@el-danihasbiarta1200]

yeah we need people like him that time

[@SWIRFTV]

i barely understand most of your videos, but i figure its like a second language, if i watch them alot even while i sleep i will absorb the knowledge. is this like theoretical physics? where measuring something actually effects the actual thing you are trying to measure?

[@Muratcharms]

So the llc to extreme or whatever setting on the bios. On actual die it can only go as high as same and "cant be more than" what u set for vcore? But its still valid you gonna have way more overshoot on load release and thats the dangerous part?

[@jeffreybouman2110]

Maybe a stupid question but how much droop of the voltage is normal? I know that on intel systems the voltage goes up under load and with AMD Zen it goes down under load but is there a sort of number you need to focus on getting up or down? Thank you very much for the great content :)

[@george_moussa]

Apologies for my lack of knowledge regarding the topic of voltage regulation and monitoring, I’m kinda new. Does CPUID HWMonitor report the VRM Controller voltage (Die sense) or the Super IO voltage? Or both?

[@gmacv3777]

HwMonitor is one of the worst one to track. I'd recommend use HwInfo64 instead.

[@resune]

Can you figure out what your loadline is by using your super io readouts as a sort of "offset"? in my z170a gaming m7 I don't have any switching freq settings or LLC settings and observe a 1.304v at idle (15-20W) and 1.28v at load (127W) with a 1.3vcore set in bios, my VID tables are a bit wonky which I'm assuming is throwing off my wattage readings (displays 1.189-1.2v), but is it still possible?

[@realjayjobes1849]

So is the video from 2 years ago wrong?

[@JasperSkallow]

Does extreme hurt things?

[@jamhough22]

So, on HWinfo you have Vcore reading and also Core X VID, are these from diferent places because they are always reading diferent? I have a maximus x CODE and ive always been very impresed with the way it handles voltages, it always seems to be solid even at stock switching frequencys, i can undervolt things nicely without much issues. I have had to set my LLC to level 5 but thatll be because im running a i7-8700K all cores locked at 4.7GHz @1.26v which isnt amazing but im happy with it.

[@Fix_It_Again_Tony]

I think the VID value in HWinfo is a request from the processor. This is important when you have a processor with turbo features. When the processor wants to up the clock speed, it requests more voltage until some limit is reached (thermal, Power dissipation, utilization, etc.). When it wants to lower the clock speed and save power it requests less voltage. The Vcore value is a measurement of the voltage in real time, from either the super IO or the VRM itself as described by the video.

[@LecoSilvaPoesia]

@@Fix_It_Again_Tony that's a grear answer! I've found researching about VCORE precisely that: VID is what the processor requests, not what it's actually getting.

[@b747xx]

So Instead of putting 1.35V into my 3950X on IDLE I should put it to 1.22V and put the LLC to maximum instead of counting on the vdrop to bring the voltage down on load, right? I heard so many time the post about being real careful with that.. But it it's like you say I just put that to extreme and problem solved... Nice.

[@elemental3656]

Uhhh if you followed the video, don't use 0% LLC trim (maximum LLC) is the way to go. You want some vdroop (watch the previous video), as that gives better transients. So experiment with Medium-High LLC. For the 3950X "manual OC voltage", I would watch an earlier video bz posted on overclocking the 3950X. Ryzen 3000 series has tons of quirks for manual OC.

[@XenHat]

this ammendum is brought to you by a defective tablet driver

[@zkkzkk32312]

Is super IO only a Intel thing? if SVI2 on Ryzen is actual VRM measurements then in HWInfo64 that number is accurate?

[@maxmustermann7397]

Nice video but I miss an all-embracing explanation here, like in another video you said that there is an overshoot from the loaded inductors after the switching too. But I still agree with Andrew. For me the information in this video don't dissent with the comment at all, the opposite is the case. Like on any electrical system there is a voltage drop if you put load to it. How big the drop is depends on the charakteristik (design) of the power source and the resistance of the consumer. So you can't effectivly change the resistance of the CPU on load or change the design of the circuit, but the VRM can add more voltage on the source to compensate the drop, so that the effective voltage on the CPU is still good. This behaviour is exacly what you have mentioned in your other videos, the closer your probe is to the CPU the lower the voltage is that you had measured (due to the line of capacitors). But because the VRM adds more voltage, it's higher if there isn't load anymore and that causes the overshoot. So it takes a while (transient response) until the VRM detects that and lowers the voltage again to the base line. Or in other words, if the voltage would be always the same there would be no under- or overshoot but then there would be no calibration for the load line anymore and so the CPU would be very unstable due to high drops on high load. This also explains why a short inconsistent load causes more stability problems than a constand load because the difference from a no operation instruction followed by a complex instruction causes a higher drop and so needs a higher voltage compensation than from a simple followed by a complex one. So the under- and overshoot is worse for the first and can cause more stability issues compared to the latter even the CPU needs less power then.

[@rankcolour8780]

Transient under/overshoot is independent of LLC. It's more to do with VRM response times than electrical resistance.

[@maxmustermann7397]

It seems to me that you haven't understand me because I don't wrote what you have implied. The VRM regulates the voltage to a choosen value and how strong this regulation is depends on the LCC settings. Meaning how much the base line differs between load and no load. So yes this alone has nothing to do with under- or overshoot and I never wrote something differently. So because this is a regulation it takes a little bit time until the new values are reached, because it's not possible to instandly change and go to a specific new output value, the time which is needed for this is called transient response time. It's impossible to jump from 2 V directly to 100 V, the value will raise from 2 until 100 V with all the values in between, but without the right equipment you aren't even able to see it because it happens very fast. The CPU has a different electrical resistance depending on the load, so it's not consistant and I had just refered that this causes the general voltage drop. What I have expressed with the text is that it's only possible to raise the base line on load by applying more voltage to the CPU. So the undershoot comes because the VRM needs time to add more voltage to the system and the overshoot because then there is more voltage applied so it takes time for the VRM to lower the voltage again. Because I agree with the statement that the whole regulation is handled by adjusting the output voltage. Or in other words, if there is no regulation at all (LCC completly off, but I don't know if that is even possible) there wouldn't be an under- or overshoot at all and instead just a fluctuating drop.

[@rankcolour8780]

@@maxmustermann7397 You would still get overshoot. Overshoot is not caused by LLC. It just is made worse by very high levels of LLC If you turn LLC off ( maximum LLC/flat voltage) Overshoot gets worse, as does undershoot. LLC is just a level of resistance that the VRM is simulating.

[@maxmustermann7397]

@@rankcolour8780 There are multiple things involved and this disccusion would get unintelligible very quickly when we start to include more things here than just LCC. So please don't let us talk about overshoot from the inductors or maybe other things. But despite from that if overshoot isn't caused by LCC what else causes it? How are you sure that the overshoot from the LCC is related to it and not independet from it? If the regulation isn't done by adjusting the voltage how else? In addition please don't mix up words and stick to one contemplation. You are using terms in a weird conjunction, calling it off and maximum at the same time isn't reasonable. What I'm saying from the beginning is if the LCC settings are set to high, the base voltage line which the CPU sees is the same / flat overall regardless of the load, meaning the average voltage if there is load and no load is nearly the same. So I say to archive this the VRM needs to add more voltage to the system, like to apply 1,1 instead of 1,0 on the source side so the CPU still get 1,0 on the other end of the wire. But when were isn't a load anymore the CPU will get the full 1,1 for a short time until the VRM detects it and reduces the voltage back to 1,0 again.

[@rankcolour8780]

@@maxmustermann7397 maximum llc is zero load line calibration. You seem to have your terms mixed up. As load line level decreases resistance and droop increases, increased loadline is less droop because Load line is not voltage regulation it is resistance based on current. If you turn off all LLC you would have flat voltage and lots of overshoot and undershoot. The reason you are missing an all embracing solution is because transient overshoot and undershoot and load line are not the same thing so they cannot be explained with one notion. If you stop telling yourself LLC is voltage regulation and call it what it is levels of programmable resistance you will understand both. Voltage regulation is a result of VRM switching, not loadline. Load line is resistance VOltage regulation is VRM switching Over/undershoot is the inability of the VRM to keep up with changing CPU power states combined with the function of caps. Zero LLC results in the largest transients, large amounts of LLC (resistance)result in the lowest

[@ElbruzMafe]

I really wonder something that I have big difference my vcore values for each ones for example 0.9 to 1.4. Is it normal for overclocking ?

[@Krzeszny95]

ASRock, at least Extreme4 has it backwards. Level 1 is 0% and Level 4 is 100%. UEFI says Auto is Level 1/disabled. I set it to Level 2 and hopefully it's ok.

[@mylittlepwny3447]

So on an ASUS board is leaving LLC at Auto good? It seems super solid to me

[@jake20479]

can you do a short video about VR Vout?

[@LecoSilvaPoesia]

I've been trusting the VR VOUT sensor since this: www.overclock.net/forum/27736104-post17.html

[@TillTheLightTakesUs]

Please answer me this: I'm on a 9700k@4,9ghz with maximus xi hero and I have Lev 3 LLC with an OFFSET voltage of +0,005v. The CPU draws around 1.19-1.252v depending on how high the load is, and it doesn't BSOD at all. Now here are my questions: Can I use this config forever without any future harm? And if yes, can I OC further, and if yes, how high can I go? Secondly, is it better to use a high LLC with low Vcore, or vice versa (sticking to intel spec LLC)? Thanks.

[@randomchars9619]

not sure if that video reduces confusion, b/c its halfbaked. now ppl may think theres no overshoot anymore and that they cant damage their cpu on the long run with an extreme setting (concluding from comments) whats not the case. the only thing changed is that llc does not (OVER)compensate, just taeks stuff away from what is already there. that doesnt make transient response peaks less dangerous. so maybe make an all-in-one llc vid, with updated knowledge, that covers best practices.

[@spamlucal]

Does that means that if the LL register in the voltage controller has less mOhms programmed than the actual hardware, it has the same effect as increasing LLC in BIOS? (apart from all other side effects)

[@ActuallyHardcoreOverclocking]

yes setting the LL register to 1mOhm with 100% in the LL trim register gives the same Vdroop as 2mOhm LL with 50% LL trim.

[@sagargupta9014]

Does this means I can skip touching LLC and increase vcore

[@personaldronerepair6141]

I always walk away smarter.

[@robsamp]

I think you will need one more video talking about Load Lines... LOL. It's "impossible" to talk about VRM, VID, VCore and load lines without talk about DC_LL and AC_LL.

[@대괴수대백과대사전]

Maybe this explains strangely high temperature with Z390 ASUS boards when OCing. Maybe ASUS voltage readouts are showing voltage from the Super IO which is quite lower than actual voltage. I've worked with quite a lot of MBs and OCing on ASUS's Z390 boards result in unusually high temperature @ same reported voltage.

[@GurkenkoenigZ]

Have you watched and listened to the vid? I think someone is facepalming to your comment. Afaik as i understand now, Asus's higher end Boards copy the die-sense. And then when you read a cheaper one it doesnt. So if both Boards report the same voltage, the cheaper one is actually less voltage. And that less voltage is why less powerdraw/heat. And in case i misunderstood, hit me daddy.

[@rankcolour8780]

That's more likely just down to crappy heatsink design combined with less VRM surface area

[@Sophistry0001]

How would one know where the Vcore measurement is taken on their specific motherboard? I didn't realize ASUS boards were that bad, I guess that's 300 bucks down the drain for the ROG x570-e. I would have thought that all boards in that price range and above have all the shiniest bells and whistles. I guess turning up LLC on my Ryzen 3950x won't help me actually hit single core boost clocks as printed on the box, but will it help increase my all core boost?

[@kelvinyonger8885]

Depends on Vmin, which you need to solder to the board and use an oscilloscope to get. If you're fine with noise/high temps the highest LLC is fine.

[@TheBLC94]

@@kelvinyonger8885 and higher voltage peaks that will slowly degrade your chip, most importantly

[@jestt600]

Thanks, However I am still not sure what setting I should go for!! I got an 9900k running @4.7 on all cores on a z390 Aorus pro wifi board, 32GB memory running @3200. Do I keep it at default or should I set it for high?

[@nhozdien5058]

jestt600 if you never overclock before, just do some researches and have some fun.

[@davidjohansson1416]

what's up with 11% likes? Buildzoid deserves more!

[@OlettaLiano]

People are too lazy to click a like button.

[@levyroth]

that's the average for all social media algorithms. even on IG the average of likes is around 10% and higher, obviously, if you have a viral post. it's always been like this.

[@ilkanakten]

How am I going to know that my ASUS board has this fix or not? What is the name of it, digi+ vrm? My board is tuf gaming z490 WiFi

[@jboy1757]

So my 5 ghz OC is stable but when I play a game for approx 20 minutes, the game “dies”. I can immediately start the game back up but approximately 10 minutes later, it “dies” again. My LLC is set to 4. Will setting the LLC to 5 help or hurt with stabilizing the OC during gaming?

[@LecoSilvaPoesia]

I'd recommend you to check VR VOUT sensors. I usually run Geekbench 64-bit to check if the manual VCORE I've set + LLC is good for the load I'm undertaking. What happened to me was that when the VCORE + LLC got me down to a voltage insuficcient for the OC, it'd either crash with a nasty BSOD or soft lock and then close the test.

[@CUAUHTEMOCVEGATORRES]

27:17 but why? your issue is that you dont give a direct and concrete answer, please dont workaround, be straight and direct, that tends to make this video so tedious and is not.

[@Pedro-tl7jg]

Wtf my llc settings are from lvl 1 to 6 what is a medium setting here? 3? and high would be 4 5 6? pls help

[@soraaoixxthebluesky]

So the solution is all board manufactures should replicate a true VRM controller voltage so we consumers don’t get confuse

[@butre.]

so should I just be cranking my llc and lowering my voltage then?

[@ActuallyHardcoreOverclocking]

watch the previous video for why Load line exists.

[@noe8615]

@@ActuallyHardcoreOverclocking cant you just simply answer the question ?

[@ertai222]

@@noe8615 apparently not

[@maxspamer8744]

Ну так в Итоге как правильно то???? на low или medium ставить???

[@Z1mmb0]

You are freaking funny :D

[@Michael-OBrien]

Those inputs for the controller and the SuperIO chip should be high impedance. Why isn’t the SuperIO sampling from it? If anything, they can buffer the input on the SuperIO… Are you using x10 probes? The capacitive loading of the probe is a real thing that’s creative your issue. If you spent some [Patreon] money on a good probe, you’ll be set. Or buy a differential probe, but those are $$$$.

[@keubis2132]

31:50 funny story ryzen master bricks my bios and i need to remove the battery to completly reset it (dunno if bricks is the proper word tho)

[@tjsmithson1598]

It's not. Bricked would imply the vBios chip is corrupt or something.

[@danielson_9211]

New to OCing and im like um um deer headlight, have an Asus with a 10700k set to AI everything and im getting 5.2ghx turbo and 4.9 with avx 1.325V 76C temp prime95 , I changed it to sync all cores and not a chance in hell stable but boots LOL. tried to using LLC, uping the voltage to 1.35 cant make it stable but the dam AI can. Trying to use photoshop or play games and I can see the cores bounce up and down like a yo yo haha. If you could assist it would be very appreciated.

[@soundspark]

Had to run LLC medium on my Gigabyte motherboard just to have my 13900K stable at stock speeds.

[@TheMakiran]

Well, I still didn't get why in this case extreme LLC is bad. Can someone please explain me?

[@kelvinyonger8885]

The min voltage is still the same, so it's inefficient.

[@stoneymahoney9106]

Vdroop is a thing because the CPU core and VRM are a pair of resistors in series, otherwise known as a voltage divider, and the way the resistances change as the output voltage and load change cause the VRM resistance to increase at a proportionally higher value than the CPU. Trying to completely remove Vdroop is effectively engineering away one of the basic laws of physics as applied to electronics, the result of which is most notably larger load-off voltage transients. You're exchanging a slightly lower voltage under load for rather nasty voltage spikes every time the CPU stops doing something. The compromise of a little bit of Vdroop and far more managable transients is the easiest, cheapest way to make the situation acceptable. EDIT: See Buildzoid's recent follow-up video - modern VRM designs mean the load-off transient overshoots don't increase a lot, but the undershoots that immediately follow result in a short period of reduced Vcore where a quick reapplication of load to the CPU is much more likely to cause a crash (srsly, watch the video, the details are important, it's a balancing act)

[@brovid-19]

I'm more angry he called it "IxR" losses. _It's I² x R losses_ If you're gonna be wrong, at least be wrong using correct math. sheesh.