Jump to content

Load-Line Calibration: why overclockers should care

 

WAN show featured! :D

 

LLC no longer affects vcore stability in Haswell or Devil's Canyon processors, as all the voltage regulation happens on the cpu.  LLC only affects Vin (the voltage supplied to the cpu as a whole), and as far as I can tell, Vin stability has no effect on Vcore stability.  More information can be found 

in this thread.  

 

First a disclaimer: this is my experience with Load-Line Calibration (LLC) during my overclocking.  Some of the information in the post may be incorrect, although I will try to only post information which I feel is validated given my experience.  I don't have any certification in making motherboards or programming, so this is simply my understanding of how this stuff works.  

Please remember that every motherboard manufacturer may use a different bios setting to implement LLC, so make sure that you look at what YOUR motherboard/manufacturer says about their LLC implementation.  

 

 

Why I'm writing this:

This subject has been briefly addressed by several people on several different forums and several different overclocking guides. But a quick google search for load line calibration gave an article from 2010 (http://www.overclockers.com/load-line-calibration/) concluding that LLC was good and overclockers should use it.  Other threads that come up with a google search are usually two-liners asking whether they should use LLC for their overclock, with the general consensus being yes they should.  I couldn't find anything about LLC on the LinusTechTips forum, and I felt that this should be addressed.  

 

There seem to be no recent go-to beginner-level threads about LLC, and there seems to be no general understanding of what it is actually doing.  This thread is intended to be an introduction to what LLC is actually doing, and why you should use it with care.  

 

 

Background on LLC:

For those of you who don't really know what LLC is: LLC was a featured added to motherboards several generations ago to combat vdroop.  Vdroop is a drop in voltage supplied to the CPU as load increases; basically when you go from idle to load, the voltage would decrease.  Given the small voltage tolerance that overclockers are working with (increased voltage is proportional to the CPU frequency/multiplier that an overclock can achieve), a droop in voltage applied to CPU can make a theoretically stable overclock unstable (dropping the voltage below that required to achieve the set frequency).  LLC applies additional voltage to the CPU to combat vdroop so that when switching to load, there is sufficient voltage to keep that frequency stable.  So LLC is great and you want to turn it on? Yes, but...

 

For most modern motherboards, there are different levels of LLC that you can set in your bios.  At certain levels of LLC (these may be different for each motherboard), the LLC can overcompensate for this vdroop, and actually apply vboost.  Vboost is when the voltage actually supplied to the CPU is above the value that you set in your bios.  This can be a nice way of ensuring that your overclock will be stable, but you have to be careful, because each CPU has a death voltage (the voltage where, if applied to your CPU, it will likely die).  If you are toeing the line near your CPU's death voltage to try to squeeze every last MHz out of your overclock, LLC can bring your actual voltage above this level, which is a great way of killing your CPU (or making it degrade much faster).  So although LLC is great for overclockers, it should be used with care, because you may just end up killing your CPU.  

 

Now each motherboard is different, and may label their LLC settings differently, and this thread will be based on my settings on my ASUS Rampage IV Extreme.  Make sure that you check your motherboard manual (and do a bit of googling for other people's experience) for how LLC is implemented in your case.  (For instance, Asrock motherboards from the H77 generation had their LLC values reversed from the values that ASUS uses.)

 

Now I had heard of all this LLC mumbo-jumbo (actually it was thecrazyrussian who told me to be careful with my overclock), and I wanted to see just what actually happened when you try different levels of LLC.  So I went through my motherboard and tested each LLC setting and ramped up the set voltage, seeing what the actual read voltage was.  I used a digital multimeter (just a basic one that I bought at the Source), but even that is more reliable than the values that your motherboard bios (or software) can read.  

 

 

Testing Setup:

ASUS Rampage IV Extreme, i7 3930k, 16Gb Corsair Vengeance CL8 ram in quad-channel, XFX PRO 1000W PSU, Corsair H100i cooling.  The amount of Vdroop changes with the CPU frequency, so I set my multiplier at 40 (stock is 32, and I have a stable overclock for this CPU at 45).  I started at an arbitrarily chosen Vcore of 1.325V and ramped up until 1.4V (the voltage past which my CPU degrades much faster, and is considered by some to be the near death voltage of the CPU), or until the temperatures hit about 78 C.  To avoid having to reboot with every voltage change, I applied all of my voltage tweaks using ASUS' AI Suite 2.  Idle voltages were taken at the Windows 7 desktop with no programs open, load voltages were taken after Prime95 (small FFTs) completed its first pass.  My motherboard has five settings for LLC: Regular (0%), Medium (25%), High (50%), Ultra High (75%), and Extreme (100%).  My Vcore (CPU voltage) can be changed by steps of 0.005V, which may seem very small, but keep in mind that a change of 0.005V in Vcore can destabilize an overclock.  

 

Results

LLC_Regular.png

At a LLC setting of Regular (0%), the voltage at idle was an average of -0.018V from set (idle: blue line, set: black line), the voltage at load dropped an average of -0.054V from set (red line), with a droop from idle to load of -0.036V.  Immediately we can see that not enabling LLC can seriously destabilize an overclock.  

 

LLC_Medium.png

Moving up to an LLC setting of Medium (25%), the average voltage changes from set were -0.007V at idle, -0.023V at load, with an idle to load droop of -0.016V. The idle voltage isn't that bad, being pretty close to the set voltage, but the load droop is still more than enough to destabilize an overclock.  

 

LLC_High.png

Now up to an LLC of High (50%), the average voltage changes from set were +0.005V at idle, +0.011V at load, with an idle to load boost of +0.006V.  This LLC appears to be pretty good, with the motherboard actually putting out a similar voltage to the one we set in the bios.  There is a small amount of Vboost, but the magnitude is unconcerning, putting us nowhere near the death voltage of the CPU.  For this LLC setting I only went to a setting of 1.380V, because CPU temperatures were becoming concerning.  

 

LLC_UltraHigh.png

This is where things start to get interesting.  Setting a LLC of Ultra High (75%) gave average voltage changes from set of +0.018V at idle, +0.045V at load, with an idle to load boost of +0.028V.  This is hugely different from what we set in the bios, idling at ~3x and loading at ~9x the increments we can increase and decrease by in the bios.  Here I stopped increasing values for two reasons: the first being that the CPU was at 77 C, and the second being that the actual read voltage was just barely below the fast-death voltage for my CPU.  

 

LLC_Extreme.png

I was ready to stop at Ultra High, but to do my due diligence, I tried Extreme (100%) LLC.  The idle voltage was +0.031V above set, and the load voltage was an insane +0.086V above set.  Just switching it to load brought the voltage well above my 1.4V ceiling.  I didn't even let my prime95 get to the first pass, I just took the reading and brought the computer down as fast as I possibly could.  

 

 

Conclusions

Quite frankly I was shocked to see the effect that LLC setting has on actual voltages, especially at Ultra High and Extreme.  I do understand that that every motherboard may implement LLC differently, and the Vdroop/Vboost changes may not be as incredible as I saw on my board.  I can easily visualize someone trying to get the highest overclock possible, but ignoring the LLC setting (or worse setting it to extreme) and frying their CPU.  I hope this thread illustrates my experience with LLC and persuades the reader that LLC should be used when overclocking, but must be used with care.  

 

Personally I chose an LLC setting of High (50%) for my overclocking, because it resulted in no Vdroop, but didn't result in enormous Vboosts.  I also took into account the small observed Vboost, and made sure to never bring my voltage to a level where the Vboost would touch the fast-death voltage of my CPU.  I have what I consider to be a stable overclock with this motherboard and CPU at 4.5 GHz at a Vcore of 1.325V (stable for 24h of prime95 small FFTs).  

 

Note to the reader after additional testing:

Vdroop and Vboost will not behave in a fixed manner!  Idle and load voltages follow a linear trend, but the slopes of those lines are not equal.  Read my follow up post for more details.  

 

TL;DR:

LLC should be used while overclocking, but used with care.  If you don't and you're not careful, you could kill your CPU or degrade it very quickly under load voltages.  It can also be chosen logically, see part 2 for more details.  

 

Read Part 2 here!

 

Isopropyl alcohol is all you need for cleaning CPU's and motherboard components.  No, you don't need [insert cleaning solution here].  -Source: PhD Student, Chemistry


Why overclockers should understand Load-Line Calibration.


ASUS Rampage IV Black Edition || i7 3930k @ 4.5 GHz || 32 GB Corsair Vengeance CL8 || ASUS GTX 780 DCuII || ASUS Xonar Essence STX || XFX PRO 1000W

Link to comment
Share on other sites

Link to post
Share on other sites

I'm using LLC on my 3570K currently.

"It pays to keep an open mind, but not so open your brain falls out." - Carl Sagan.

"I can explain it to you, but I can't understand it for you" - Edward I. Koch

Link to comment
Share on other sites

Link to post
Share on other sites

LLC was easily the most annoying thing about overclocking my 2600K.

 

I found out about vDroop when I first built my PC and I couldn't get past 4.2 on any voltage.  I'm lucky I didn't fry my CPU setting it to 1.4-something just to get 4.2 stable...

 

And after a buttload of Google searching I finally came across one graph on the whole Internet that described how LLC works on my Gigabyte mobo.  Found it on some random forum somewhere.

 

It's still annoying that the value set in the BIOS never matches the actual voltage, but at least I can now get my Vcore pretty constant (it's easy enough to monitor with HWMonitor).  Out of 10 LLC levels on my mobo, Level 7 gives almost no idle to load boost, which is nice.  The voltage does jump around a bunch at idle, though, which is still annoying, but whatever.

 

Anyway, there's my vDroop rant.  I'm glad it's gone with Haswell.  An advantage of having VRMs on the CPU die I suppose.

Intel Core i7-7700K | EVGA GeForce GTX 1080 FTW | ASUS ROG Strix Z270G Gaming | 32GB G-Skill TridentZ RGB DDR4-3200 | Corsair AX860i

Cooler Master MasterCase Pro 3 Samsung 950 Pro 256GB | Samsung 850 Evo 1TB | EKWB Custom Loop | Noctua NF-F12(x4)/NF-A14 LTT Special Edition

Dell S2716DGR | Corsair K95 RGB Platinum (Cherry MX Brown) | Logitech G502 Proteus Spectrum | FiiO E17 DAC/Amp | Beyerdynamic DT990 Pro

Link to comment
Share on other sites

Link to post
Share on other sites

Yeah. My voltage right now is technically set to 1.28, but LLC is on Ultra High currently so it's at around 1.272. At 4.4GHz.

"It pays to keep an open mind, but not so open your brain falls out." - Carl Sagan.

"I can explain it to you, but I can't understand it for you" - Edward I. Koch

Link to comment
Share on other sites

Link to post
Share on other sites

  • 1 month later...

Yeah. My voltage right now is technically set to 1.28, but LLC is on Ultra High currently so it's at around 1.272. At 4.4GHz.

 

Is that under load? That seems like a small jump for Ultra High to me, although I guess every motherboard is different.  

Isopropyl alcohol is all you need for cleaning CPU's and motherboard components.  No, you don't need [insert cleaning solution here].  -Source: PhD Student, Chemistry


Why overclockers should understand Load-Line Calibration.


ASUS Rampage IV Black Edition || i7 3930k @ 4.5 GHz || 32 GB Corsair Vengeance CL8 || ASUS GTX 780 DCuII || ASUS Xonar Essence STX || XFX PRO 1000W

Link to comment
Share on other sites

Link to post
Share on other sites

On my gigabyte board i set my voltage at 1.212v for (3770k) 4.3ghz,

 

I have LLC set to extreme, and the voltage goes upto 1.236v.

 

That is with C3/C6 enabled, with it disabled. it only goes to 1.224v

 

PC is rock solid 24/7 stable.

----Ryzen R9 5900X----X570 Aorus elite----Vetroo V5----240GB Kingston HyperX 3k----Samsung 250GB EVO840----512GB Kingston Nvme----3TB Seagate----4TB Western Digital Green----8TB Seagate----32GB Patriot Viper 4 3200Mhz CL 16 ----Power Color Red dragon 5700XT----Fractal Design R4 Black Pearl ----Corsair RM850w----

Link to comment
Share on other sites

Link to post
Share on other sites

Is that under load? That seems like a small jump for Ultra High to me, although I guess every motherboard is different.

Yes. That's under load. Though I'm using a different board now and have been able to achieve lower voltages on the same frequency.

"It pays to keep an open mind, but not so open your brain falls out." - Carl Sagan.

"I can explain it to you, but I can't understand it for you" - Edward I. Koch

Link to comment
Share on other sites

Link to post
Share on other sites

I made a thread, but didn't quite get the answer I wanted yet, but this is quite relevant to it. I OC'ed my cpu to 4.2 ghz (2500k) on an asrock extreme3 gen3 motherboard, and I set the voltage to manual at 1.255 volts. I got to the desktop, everything is stable, but my voltages aren't at all what they should be. At idle it's 1.2 volts, and load it's between 1.13 and 1.15. What could cause such a massive vdroop effect? My PSU is a corsair AX860i platinum. 

Link to comment
Share on other sites

Link to post
Share on other sites

I thought everyone is using LLC for CPUs for overclocking. :huh: Nice post.

A water-cooled mid-tier gaming PC.

Link to comment
Share on other sites

Link to post
Share on other sites

I thought everyone is using LLC for CPUs for overclocking. :huh: Nice post.

It should only be used if you suffer from Vdroop, which isn't everyone. 

"It pays to keep an open mind, but not so open your brain falls out." - Carl Sagan.

"I can explain it to you, but I can't understand it for you" - Edward I. Koch

Link to comment
Share on other sites

Link to post
Share on other sites

Nice write up. Now I know where to send people when their overclocks are failing.

I don't think you mentioned anything about offset vcore and how offset has it's own LLC. Sooo, when attempting an offset overclock, leave LLC disabled or at 0%. That'll allow you to have one core at (100%(ex:)) x49 and 1.4v for single threaded tasks. Then when you're under 100% full load and the vdroop comes in... You've already set your other cores to hit x48 and your vcore droops to a sane, stable 1.35v.

yee. makes since, right? works too.

Link to comment
Share on other sites

Link to post
Share on other sites

It should only be used if you suffer from Vdroop, which isn't everyone. 

Ok.

A water-cooled mid-tier gaming PC.

Link to comment
Share on other sites

Link to post
Share on other sites

I don't think you mentioned anything about offset vcore and how offset has it's own LLC. Sooo, when attempting an offset overclock, leave LLC disabled or at 0%. That'll allow you to have one core at (100%(ex:)) x49 and 1.4v for single threaded tasks. Then when you're under 100% full load and the vdroop comes in... You've already set your other cores to hit x48 and your vcore droops to a sane, stable 1.35v.

 

Yeah I didn't mention anything about offset as I don't use it, personally I want to set it to one voltage and have it stay there.  I'm not a fan of major voltage fluctuations.  

 

An appropriate LLC setting should be worked out by each overclocker, to their preference.  It seems to me that you could still use LLC and offset in conjunction, you would just have to be SUPER careful.  

 

Your theory is sound, provided that the maximum vdroop still provides enough for the OC to be stable.  

Isopropyl alcohol is all you need for cleaning CPU's and motherboard components.  No, you don't need [insert cleaning solution here].  -Source: PhD Student, Chemistry


Why overclockers should understand Load-Line Calibration.


ASUS Rampage IV Black Edition || i7 3930k @ 4.5 GHz || 32 GB Corsair Vengeance CL8 || ASUS GTX 780 DCuII || ASUS Xonar Essence STX || XFX PRO 1000W

Link to comment
Share on other sites

Link to post
Share on other sites

I OC'ed my cpu to 4.2 ghz (2500k) on an asrock extreme3 gen3 motherboard, and I set the voltage to manual at 1.255 volts. I got to the desktop, everything is stable, but my voltages aren't at all what they should be. At idle it's 1.2 volts, and load it's between 1.13 and 1.15. What could cause such a massive vdroop effect? My PSU is a corsair AX860i platinum. 

 

The PSU is likely not your problem, you probably don't have your LLC set to an optimum setting, try increasing it one step and see what your vdroop is like going from idle to load.  

 

You may also have sleep states enabled, which will lower your voltages and multipliers when in idle.  

Isopropyl alcohol is all you need for cleaning CPU's and motherboard components.  No, you don't need [insert cleaning solution here].  -Source: PhD Student, Chemistry


Why overclockers should understand Load-Line Calibration.


ASUS Rampage IV Black Edition || i7 3930k @ 4.5 GHz || 32 GB Corsair Vengeance CL8 || ASUS GTX 780 DCuII || ASUS Xonar Essence STX || XFX PRO 1000W

Link to comment
Share on other sites

Link to post
Share on other sites

I use LCC on very high to mt FX8120.

FX8120 @ 5.0GHz | Sabertooth 990FX | Patriot Viper Xtreme 8GB @ 2133MHz | Powercolor HD7950 @ 1300/1800 | Bitfenix Shinobi XL Modded | Corsair AX850 | Mushkin 120GB & Seagate 2TB | Bitfenix Recon | 5X SP120 PE | 6X Yate Loon 120mm HS


Koolance 200 Res | 2X Koolance 360 Rads | Danger Den CPX Pro | EK 7950 Block | Koolance CPU Block | Bitspower & XSPC Fittings


Green Machine

Link to comment
Share on other sites

Link to post
Share on other sites

excellent post!!  Helps me figure out some of my problems!

My build:  Leviathan  Case: 900D  CPU: i7 3770K (watercooled)  Mobo: Z77X-UD5H GPU: EVGA GTX 780 Hydro Copper GPU: MSI GTX 780 watercooled PSU: EVGA 1300W G2  RAM: 32 GB Corsair Vengance  HDDs: 1 x 120 GB Intel 330 SSD (OS X); 1 x 256 GB Samsung 840 pro (Windows 8); 2 x 2TB Seagate Barracuda (RAID 0 Data OS X); 1 x 3TB Seagate Barracuda (OS X backups)  Monitors: 1 x 24" Apple LED Cinema (center); 2 x 23" Apple LED Cinema (surround)  Watercooling: 3 rads, CPU, GPU, GPU, MCP655 pump, Lots of fittings, EK reservoir, EK UV Blue coolant.  Updated build: Leviathan 2.0

Link to comment
Share on other sites

Link to post
Share on other sites

  • 2 months later...

I'd like to note that this varies from board to board. It's similar to how Z77 Asrock boards dump way more voltage into the chip than CPU-z says; it's just a design flaw in the board.

 

post-1491-0-55232400-1381070677.jpg

 

post-1491-0-23704700-1381070686.jpg

Link to comment
Share on other sites

Link to post
Share on other sites

Apparently, LLC going way above set voltages is common on ASUS boards especially on those that are advertised for overclocking.

~meOw! Σ:3

Link to comment
Share on other sites

Link to post
Share on other sites

It looks like ASUS is implementing LLC in their new ROG Matrix video cards (at least it's on the R9-280x).  We'll have to see how they implement it and whether it will be as dangerous as it can be on their ROG boards.  

Isopropyl alcohol is all you need for cleaning CPU's and motherboard components.  No, you don't need [insert cleaning solution here].  -Source: PhD Student, Chemistry


Why overclockers should understand Load-Line Calibration.


ASUS Rampage IV Black Edition || i7 3930k @ 4.5 GHz || 32 GB Corsair Vengeance CL8 || ASUS GTX 780 DCuII || ASUS Xonar Essence STX || XFX PRO 1000W

Link to comment
Share on other sites

Link to post
Share on other sites

im usin high llc atm, with it on auto I crash during stress tests 

Stuff:  i7 7700k @ (dat nibba succ) | ASRock Z170M OC Formula | G.Skill TridentZ 3600 c16 | EKWB 1080 @ 2100 mhz  |  Acer X34 Predator | R4 | EVGA 1000 P2 | 1080mm Radiator Custom Loop | HD800 + Audio-GD NFB-11 | 850 Evo 1TB | 840 Pro 256GB | 3TB WD Blue | 2TB Barracuda

Hwbot: http://hwbot.org/user/lays/ 

FireStrike 980 ti @ 1800 Mhz http://hwbot.org/submission/3183338 http://www.3dmark.com/3dm/11574089

Link to comment
Share on other sites

Link to post
Share on other sites

  • 3 months later...

Load-Line Calibration (LLC) Behavior Part 2: now with more science! (Read part 1 here)

Necroed with permission from Windspeed36

 

So as fun as it was to be featured on the WAN show, I took particular offense to Linus saying that it wasn’t the most scientific thing ever.  As a BSc (Chemistry) and a current MSc student, it was now a point of honor to do a better job.  Luckily, I got a new motherboard, so I have an excuse to try again!

 

In my previous post/article/report (whatever you want to call it), I outlined how LLC behaves at different settings while at idle and load.  I drew attention to the fact that higher values of LLC will result in a massive voltage jump going from idle to load.  In this sense I think I can agree with Linus’ “lack of science” statement, where I just didn’t record enough readings.  So here I am trying again, expanding the testing range, and performing statistical analyses on the data collected. 

Before I get too into the nitty-gritty details, I do want to briefly touch on why this is important.  When you pull your motherboard out of the box, LLC will (in all likelihood) be set to AUTO.  The unfortunate thing about the AUTO setting for overclocking, is that you don’t know what it will do.  Frequently the AUTO setting will dynamically choose a setting based on what you’re doing.  This is why people always recommend doing your own overclocking as opposed to clicking the “automatically overclock” button.

Imagine, if you will, yourself trying to get an overclock with AUTO selected for LLC.  What if your motherboard chooses Ultra High or Extreme as the auto setting?  How much extra voltage is it applying under load? Is that voltage degrading your CPU more than you think it is?  These questions illustrate why manually setting your LLC is important in overclocking. 

 

Note to the reader: if you don't care how I did it, you can simply just skip to the Conclusions section to get the important information.  

 

Testing Setup:

ASUS Rampage IV Black Edition, i7 3930k, 16 GB Corsair Vengeance CL8 (CMZ8GX3M2A1600C8) ram in quad-channel configuration, XFX PRO 1000W PSU, Corsair H100i cooling.  Again, my multiplier was set to 40 (stock is 32), and I ran a voltage sweep from 1.250-1.400V in 0.005V increments (expanded from my sweep of 1.325-1.400V from last post).  Voltages were applied using the ASUS AI Suite 3 (I’m too lazy to reboot between each voltage change).  Voltage was first swept at idle (1.250-1.400V), then swept at load for each LLC setting.  As with my previous motherboard, the R4BE has five settings for LLC: Regular (0%), Medium (25%), High (50%), Ultra High (75%), and Extreme (100%).  Voltage readings were taken using the ProbeIt voltage read points on the motherboard, using a digital multimeter.  Idle voltages were taken while Windows 7 was running at the desktop with no programs open, and load voltages were taken with Prime95 small FFT’s running.  Voltage readings were taken whenever possible, but voltages greater than 1.410V were a stopping point for me, and much of the load sweep for Regular LLC was unstable, so readings could not be taken in some ranges. 

 

Data Analysis:

Linear regression analyses (Microsoft excel LINEST function) were performed on the obtained data to determine the trends within each sweep.  I will present the data in chart format here for illustration, but raw data and LINEST output can be viewed here

For those who haven’t taken a statistics class, a linear regression generates a line-of-best-fit, which can be used to predict values outside of the tested range, provided that the calculated line fits the data well.  Regression outputs for the Idle and Load will be presented directly below the respective legend titles.  Each output will include the equation of the regression line (in the format y={slope}x + {intercept} and can be used as the calculation {voltage read} = {slope}*{voltage set} + {intercept}).  Below each equation of the line is an “r-squared” value, which gives an indication of how well the calculated line fits the data obtained.  A high r-squared (>0.98, say) indicated that the line is a very good fit to the data. 

 

Results:

R4BE_LLC1Regular.png

At the LLC setting of Regular (0%), we again see an idle voltage lower than set, although not as much of a drop as it was on the RIVE.  With only looking at the output of this chart we see that the lines look pretty straight, and there is a reasonable droop going from idle to load.  The interesting part of this chart comes from looking at the linear regression output. 

One would expect the read voltage to increase in a linear fashion with set voltage with a 1:1 ratio.  For each incremental increase of set voltage, I would expect the read voltage to increase by that same increment.  That is to say, I would expect the slope of each of these lines to have a value of 1.  This is not the case.  Looking closely at the data, both the idle and load slopes are not equal to one, and in fact are less than one, meaning that each unit increase does not add the full 5mV to the read voltage.  Also, the slope decreases going from idle to load. 

 

R4BE_LLC2Medium.png

At the LLC setting of Medium (25%), the idle voltage tracks reasonably closely to the set value, but there is a voltage droop when going to load.  The slopes of these trendlines are still not equal to 1, but they are closer to 1 than the trendlines of LLC Regular. 

 

R4BE_LLC3High.png

At the LLC setting of High (50%), both the idle and load voltages are slightly higher than the set values, but are notably very similar going from idle to load.  Also, the slopes of the trendlines are now greater than one, indicating that we should expect more voltage per unit set than less at this LLC setting.  These slopes are now the most similar to 1:1, and the intercepts are the closest to zero of all the LLC settings. 

 

R4BE_LLC4UltraHigh.png

At the LLC setting of Ultra High (75%), we can now see an appreciable boost in voltage from set at both idle and load.  Following the emerging trend, the slopes of the trendlines are increasing with increasing LLC setting, and the slopes are now decidedly greater than the expected 1. 

 

R4BE_LLC5Extreme.png

At the LLC setting of Extreme (100%), there is a significant boost in voltage from set at both idle and load.  The slope of the load line is now significantly higher than 1, adding a much greater voltage than the user set.  It should also be noted that the r-squared values are decreasing ever so slightly as compared to the other LLC settings.  This indicates that although the voltages should still be largely predictable, they are more likely to deviate from predicted by a few mV. 

 

R4BEdata.png R4BESlopetrends.png

The resulting slopes can be nicely grouped into a pretty table, and the slopes can be plotted to see that the observed trend of increasing slope with increasing LLC setting holds true.  Unfortunately the linear behaviour of the lines isn’t perfect, but we can clearly see that the load slopes increase much more significantly with LLC setting than the idle slopes. 

 

R4BEMultipliereffect.png

In the last post, I stated that the CPU voltage changes with the CPU frequency, but I wanted to back this up with some actual data.  Since modifying the base clock of the CPU is now a much less utilized overclocking technique, I instead modified the CPU multiplier.  I set a manual voltage of 1.300V at LLC High, and increased the multiplier from 12-40 while under load.  As we can see from the data, there is a slight increase of voltage with increased multiplier setting.  No doubt, this data would be much more linear in fashion if I had a multimeter that had increments of 0.1 mV instead of 1 mV.  The voltage only increases by 3 mV over the whole multiplier range, so effectively one could expect the read voltage to not drift significantly. 

 

 

Conclusions:

I can see that the assertion I made in the previous post that there are exact vdroop or vboost values for each LLC setting is incorrect.  Instead, it may be better to say that one can predict the effective (read) voltage at a particular LLC value and set voltage.  Now I did not test the entire range of voltages (this would not be practical), but it seems that over the useful voltage range of CPU vcore, the idle and load voltages follow a linear trend.  This means that the behaviour of the CPU voltage can be exactly known and combinations of LLC and Vcore can be logically chosen to achieve an effective voltage to stabilize an overclock. 

For instance, to achieve a multiplier of 45 on my CPU, I need 1.325V effective vcore.  This means, using the equations of the lines obtained from my analysis, I can solve for the required vcore setting at each LLC value.  In this example, to achieve a 1.325V load vcore, I would need to set 1.382, 1.348, 1.314, 1.265, or 1.258V at LLC Regular, Medium, High, Ultra High, and Extreme respectively.  Of course, I would need to validate whether the idle voltages would be sufficient to stabilize the multiplier.  I plan to test this assertion in a follow up series of tests, and I’ll likely post the results somewhere on this forum. 

How is this applicable to you, the reader? You could perform your own series of tests to determine the behaviour of your LLC on your motherboard.  The nice thing about what I’ve done here is that I’ve proven that read voltage increases linearly with set voltage with a consistent slope, so all you need to test is a minimum of three voltages on each setting to determine your LLC line behaviour.  Obviously, for increasing accuracy, more readings should be taken, but if you take three readings across your effective vcore range, you should be able to predict any voltage reading across and outside of that range at idle or load for any LLC setting you choose (to within a few mV). 

LLC is a powerful and necessary tool to consider while overclocking.  I hope that this thread illustrates that one can approach LLC in a logical fashion, as opposed to thinking of it as a magic voodoo setting.  One can logically choose an LLC setting appropriate to their overclock with only a few measurements, which should help to keep a consistent and known voltage. 

 

TL;DR:

It’s important to manually set LLC while overclocking.  Additionally, one can do a simple series of tests to determine their motherboard’s LLC behaviour at idle and load, and logically select their LLC setting to fit their overclocking needs.  

Isopropyl alcohol is all you need for cleaning CPU's and motherboard components.  No, you don't need [insert cleaning solution here].  -Source: PhD Student, Chemistry


Why overclockers should understand Load-Line Calibration.


ASUS Rampage IV Black Edition || i7 3930k @ 4.5 GHz || 32 GB Corsair Vengeance CL8 || ASUS GTX 780 DCuII || ASUS Xonar Essence STX || XFX PRO 1000W

Link to comment
Share on other sites

Link to post
Share on other sites

  • 4 months later...

@Queek

 

Thank you, thank you, thank you.  You have not received enough love for the work you put into testing this.

 

My experience with load line calibration can be a sort of warning to others.  If you want to skip straight to my experience with LLC, scroll down past the picture.

 

   I did use an automatic OC.  Specifically the Extreme Tuning feature on an Asus Z87-A Motherboard with an i5-4670k + Cooler Master 212 EVO(push/pull).  The auto-OC set my system to 4.7Ghz @ 1.275v.  For Haswell, this is an extraordinary result, probably top 20% and much better than anything I had done manually.

 

     I was ecstatic with the OC I had just got, thinking it was too good to be true, I did my due diligence and bench marked, stress tested, and gamed using this auto-OC.  Intel XTU - Pass, Prime 95 - Pass, Cinebench - Pass.  The only time when I was encountering a problem was with Battlefield 4. 

     While playing BF4 my computer would randomly freeze and a loud buzzing noise would emit from the speakers.  This "freeze" would happen regardless of what I was doing, I could be in the load menu, I could be in an intensive firefight, I could be climbing a ladder, I could just be standing still. I would not freeze when I set my system to stock.  The freeze and noise points to a hardware failure.  I only have my CPU overclocked because while I may have won the lottery with my CPU chip, my GPU cannot overclock to save its life.  Looking on the OC.net Haswell Overclocking Guide, this kind of crash is common for too aggressive of an OC, so all signs point to CPU.

     Temperatures were not an issue, because I am in the mid-60s when playing BF4.  I could not pinpoint what was causing the "crash"  I wasn't sure how to recreate the crash because it seemed to happen randomly, and recently, a little more frequently than normal.  One friend suggested that my CPU had suffered burn-in degradation as it was a fairly new CPU(purchased in February of 14').  I re-did the auto-OC, but my system jumped back to the exact same settings.  The next solution that I thought I had found was that my computer would sometimes randomly boot up into "power saving mode" which is the stick figure.  It would apply a negative offset and my voltage would be 1.250.  The game would still run, but I would still crash even at 1.250.  So what I did was double check and make sure that I put my CPU into High Performance mode(jet plane) to make sure my voltage is at the assigned 1.275v.  This still did not fix the issue.

t2MtMpe.png

 --------------------------------------------------------------------------------------------------------------------------------------------

     I was pretty stumped at this point so I had to start digging around and figure out what each and every setting is that was changed by the auto-OC.  This is how I landed on the subject of Load-Line Calibration.  This article was one of the first ones to show up on a google search.  Once I realized what load-line calibration was, I immediately thought that this is what the problem must be. 

    The auto-OC set my LLC to the maximum level, an 8 out of 8!  The vBoost that I was randomly getting must have been what caused the freeze.  If the 75% LLC level is causing a .050v boost at load, I am scared to think of what the full 100% LLC level was causing in additional voltage, certainly enough to shock the system into a freeze.

     I have since dropped my LLC to a 4/8, and I have yet to encounter a freeze.  I had previously been a strong advocate for the use of automatic OCs, because it "worked" for me.  I still am, only to a slightly lesser extent.  If you are going to use an automatic OC, make sure you change the load line calibration to something more appropriate.

"I genuinely dislike the promulgation of false information, especially to people who are asking for help selecting new parts."

Link to comment
Share on other sites

Link to post
Share on other sites

Faceman

Hey man, I'm glad I could help!  Just saving one person's overclock is enough thanks for me (why stop at one though :P).  

 

You're exactly the type of person who I was hoping to reach when I wrote this: folks who overclock without knowing to change LLC.  I didn't really know what would happen in the voltage got too high, but I guess now we know: it can cause system instability.  I recently built a system for a friend with a Z87-A and 3670k as well, and I used the auto-overclock setting, but not until after I had set the LLC to something safe in the bios (I think I used the middle setting, they didn't need a huge OC).  Auto-OC works, but be careful :)

 

Be sure to let us know in this thread if you run into any more problems (and spread the word)!

Isopropyl alcohol is all you need for cleaning CPU's and motherboard components.  No, you don't need [insert cleaning solution here].  -Source: PhD Student, Chemistry


Why overclockers should understand Load-Line Calibration.


ASUS Rampage IV Black Edition || i7 3930k @ 4.5 GHz || 32 GB Corsair Vengeance CL8 || ASUS GTX 780 DCuII || ASUS Xonar Essence STX || XFX PRO 1000W

Link to comment
Share on other sites

Link to post
Share on other sites

One thing to keep in mind is that different mobo manufactures and different voltages will effect the actual voltage you get vs. set.

 

It's a good guide to keep in mind but no one should use it as an end all.

 

For an example. On 1 on my boards if I set 1.472 volts in the bios and set LLC to Very High, no matter what load is experienced the CPU gets 1.472 volts. Interestingly enough if I leave LLC on Very high and set voltage to 1.504 volts, under load the voltage will actually fall to 1.488 volts.

 

 

So if you are overclocking and playing with LLC to always remember to see what voltage you get at idle and under load at various points.

Link to comment
Share on other sites

Link to post
Share on other sites

  • 6 months later...

Hi - informative stuff, thank you.

My experience with an i5 2500K on an Asus P8Z77-V are similar to yours.

My i5 2500K has 'average' overclocking abilities - the voltage it requires (actual) rises pretty steeply above 4.2GHz, and the P8Z77-V's VRM is (IIRC) only 8-phase, the Pro and Deluxe versions are 12 and 16 respectively, which would doubtlessly allow similar overclockng at lower vcore.

 

[ETA >> the P8Z77-V actually has a 12-phase VRM, I now know. Pro and Deluxe are both 16-phase].

One caveat with using 'high' or above LLC settings, though;

one can end up with well under 1.0V vcore at idle, which can cause lock-ups or BSOD's when the machine is, er, idling. I've experienced this myself, despite the machine being overnight stable with Prime95 (or any other stress test) at 4.4Ghz with vcore at load (again, actual) of 1.34V.

I currently use medium LLC and a +ve CPU voltage offset of 0.020V, previously I was using a negative offset of 0.010V and high LCC ('high' appears to add another 0.04V at load compared to the medium setting), but idle voltage would sometimes drop below 0.9v and I had a couple of BSOD's when Windows (7 Pro 64) was loading, or when the PC was just clocking down from a stress test.

 

cheers,

Mark.

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×