OC Needs more volt manually than with auto

[Xifios]

Hi guys,

I'm fairly new to overclocking cpus and i need your advice because i am really confused right now.
So i have been running my i7 3770k @4.2ghz for about 2 years now without any issues what so ever. However all I did back then was change my multiplier and then did some stress testing with checking my temps.
temperature wise everything was more or less fine, meaning i never noticed my temps to reaching 70 degrees while gaming.

Recently i decided i give overclocking another go and try to do it a bit more properly this time.
So what i did was follow this guide

I got to a point where i could get my 4.2ghz "stable" at 1.26V. Which seems pretty high when compared to others who can achive a higher multiplier with that amount of voltage. So I thought maybe I just got unlucky and got a chip that does not really like overclocking.

However I then tried something else just out of curiosity.I left my multiplier at 42 and changed my LLC back to auto and then did some benching.
The strange thing is that even under full load my vcore never went any higher than 1.24V at most according to HWMonitor.
Am I missing something here? Because when setting the voltage manually to 1.24 it wouldn't even let me boot into windows or at least crash while trying to log in.

Might this be somehow related to the fact that i set the LCC to extreme or is it something totally different like maybe false readings on HWMonitor or something else I need to take care of in the BIOS?

Like I said I am running a i7 3770k on a gigabyte GA-Z77X-UD5H

Thanks for your help.

 

[jfriend00]

The voltage you set for a manual voltage is not exactly what is given to the CPU. So, you are comparing apples and oranges. You are comparing what you set in the BIOS to what you see on VCore in a monitoring tool when set to an auto mode. Those are not the same things. In addition, you've changed LLC between the two scenarios which further clouds the comparison.

Instead, you need to leave LLC at the same setting for both scenarios. Then, set your manual voltage and run a stress test and write down the values you see for VCore. Then, set it to auto and do the same thing. Compare those two VCore values.

The BIOS voltage setting is not perfectly calibrated. It's like a loosely calibrated dial. If you turn it up, the CPU voltage will go up, but if you set it at 1.26V, you can't be sure the CPU is actually getting exactly 1.26V. So, you have to look at what the CPU is actually getting for a VCore in both cases and compare that. To actually make sure the CPU is getting 1.26V, you may have to set the manual voltage in the BIOS higher than that.

FYI, I would not assume that setting a fixed voltage is the "proper" way to do an overclock. On a much newer system, I've discovered that I get lower temperatures using an auto mode with an offset voltage and I can run the overclock just as high and just as stable. Also, don't assume that a higher LLC is necessarily better for a fixed CPU voltage. Most people crank up the LLC higher than they should and actually cause themselves additional headaches. Some vdroop is actually good.

 

[Mr.Scott]

Some vdroop is actually good.

When?
When voltage droops on a stable system, it starves the components of what they need to remain stable and thus puts your machine in an unstable state.
There in no such thing as good vdroop.

 

[EarthDog]

Technically, vdroop is the Intel spec... but that is correct. When overclocking you want to minimize that and use LLC.

As far as the OP - typically auto uses more voltage in favor of stability... so that's odd. I'd start over. Set defaults. Update your bios to the latest if it isnt already.set xmp... reboot then set voltage at 1.25V and see what vdroop you have. Minimize/eliminate it, then start with overclocking the CPU.

 

[Xifios]

Ok so I did change my LLC to Medium and was able to get it to 4.3ghz @1.26V now. I ran OCCT with linpack enabled for half an hour and I didn't get any crashes.
My temps however were around 80 degrees with spikes to 84. Would these be acceptable temps considering this test is pushing my system to the absolute maximum?

Also I noticed that with OCCT my system freezes for like 5 seconds every minute or so. Is this due to an unstable overclock or might this just be some problem with OCCT? because it seems like it does not really work that great to begin with considering i get some error message every time I stop a stress test.

 

[EarthDog]

84C is fine with stres testing. I don't use OCCT so not sure offhand.

 

[trents]

I tried the Linpack OCCT stress test and did not get any pauses. Did you monitor your temps during this? I was wondering about thermal throttling. OCCT has a built-in temp monitor.

Try the non linpack OCCT stress test. I think it's more demanding, at least if run on a CPU with AVX capability.

 

[wingman99]

LLC (Load line calibration) is for Load line.
With default settings there is load line because the operation of transistors is nonlinear current vs the voltage.

Load line (electronics)
"In graphical analysis of nonlinear electronic circuits, a load line is a line drawn on the characteristic curve, a graph of the current vs the voltage in a nonlinear device like a diode or transistor.
It represents the constraint put on the voltage and current in the nonlinear device by the external circuit." https://en.wikipedia.org/wiki/Load_line_(electronics)

So with load line AUTO when the PC processor is drawing less AMPs low load the voltage is higher. Then when the processor is drawing more AMPs high load the voltage is lower. That is the running physics of transistors.

 

[Xifios]

I got to 4.4ghz now. Although it was quite a significant jump from 1.26V to 1.31V. I still have a couple of questions and I hope you can answer them.

1.I have set my core voltage to 1.31 in my bios but i also have my LLC set to Medium, so technically it could go higher than those 1.31V right? Because right now it does not do that it only regulates itself down when it is not under full load which i think is actually good or is it? It never went above those 1.31V according to HWMonitor. (CPU VCORE max : 1.308)

2.When stress testing it get's to about 90 degrees which seems a bit too high so I am thinking maybe I should just go back to 4.3ghz @1.26V? During gaming however it was never reaching 70 degrees, the cores were mostly between 55 and 65 so maybe I can leave it at 1.31V considering I don't use this rig to do some extreme stuff like rendering.

3.What cpu stress testing software do you guys recommend to actually test for stability and how long do you run those? I read somewhere that the asus real bench is supposed to be closer to real world scenarios when it comes to load and such.
I already tested it for 15min and I noticed right away my temps where about 10 degrees cooler compared to OCCT so is this actually a viable tool to test things or should i use something else?

 

[wingman99]

You are safe with RealBench v2.56 and temperatures up to 90c with voltage maximum 1.4v.

 

[jfriend00]

When?
When voltage droops on a stable system, it starves the components of what they need to remain stable and thus puts your machine in an unstable state.
There in no such thing as good vdroop.

You want an appropriate amount of vdroop. Not too much. Not too little. What overclockers often think they want, a system that never droops and never goes over their target voltage either does not exist. In the real world, you have to accept either some vdroop or some over-volting, due to the inability of VRMs to perfectly respond to fast changing current requirements in the CPU. If you prevent all vdroop, you invite significant transient (can only be seen with a fast oscillosopce) over-voltage spikes at the level that could damage the CPU.

High LLC values will also really stress your VRMs.

Try reading this article on Intel Processor Power Delivery Guidelines. Here's one particular quote:

If you've ever overclocked a system, chances are that at some point or another you've had opportunity to become upset with your Vdroop "problem." Some users, confused as to why their system refuses to exactly match actual processor supply voltage to the value specified in BIOS, are quick to blame the quality their motherboard; still others find fault with the difference noted between their board's idle and full-load processor supply voltages. Actually, load line droop (Vdroop) is an inherent part of any Intel power delivery design specification and serves an important role in maintaining system stability. In most cases, comments regarding unacceptable power delivery performance are completely unfounded. To make matters worse, unjustified negative consumer perception surrounding this often misunderstood design feature eventually forced a few motherboard manufacturers to respond to enthusiasts' demands for action by adding an option in their BIOS that effectively disables this important function.

Here's another article Vdroop and Load Line Calibration: Is Vdroop Really Bad?. And, a relevant quote from that:

How vdroop got its bad name

Vdroop got a bad name when some older, low-end motherboards (with insufficient voltage regulators) had more vdroop than they probably should have. This caused stability problems when overclocking, so some overclockers made modifications (called a volt mod, vdroop mod, and vdroop pencil mod) to their motherboards to reduce vdroop. Because of this and in an effort to please overclockers, motherboard manufacturers created load line calibration to help reduce vdroop.

Load line calibration can cause your voltage to exceed what you set in the BIOS

Load line calibration goes against Intel's design and tries to keep the voltage constant at all times. However, with it turned on, we now aren't guaranteed we'll stay under the voltage that we set in the BIOS. And (as mentioned before) we're putting more stress on our motherboard's voltage regulators by trying to maintain a constant voltage.

Leave load line calibration disabled unless you really need it

If you have a decent motherboard, load-line calibration really doesn't buy you anything in terms of a higher overclock (at least it hasn't for me). It only artificially lowers the vcore that you'll have to set in your BIOS, but the CPU will still require the same amount of voltage when it's put under a load.

 

[EarthDog]

1. YOu need to see how LLC behaves. We don't know what medium does for your board.

2. 90C is as high as you want it to go during stress testing.

3. AIDA64 is what I use. You'll find each have their own which work for them.

You want an appropriate amount of vdroop. Not too much. Not too little. What overclockers often think they want, a system that never droops and never goes over their target voltage either does not exist. In the real world, you have to accept either some vdroop or some over-volting, due to the inability of VRMs to perfectly respond to fast changing current requirements in the CPU. If you prevent all vdroop, you invite significant transient (can only be seen with a fast oscillosopce) over-voltage spikes at the level that could damage the CPU.

High LLC values will also really stress your VRMs.

Try reading this article on Intel Processor Power Delivery Guidelines. Here's one particular quote:

[/COLOR]

That article is 12 years old, note.

There isn't really an appropriate amount to be honest.

I have to say that when I use LLC (if a board needs it) my goal is to set it to no droop. As in, what I set in the BIOS is what I get under load. This exists... and typically is pretty easy to accomplish on most motherboards. I've been doing this for literally over a decade across multiple generations of CPUs without damage. It puts more stress on the VRMs... making them do a better job at keeping the voltage at a more stable level.

LLC on modern boards does not lower voltage.

I think you may be following that a bit too literally.

 

[Mr.Scott]

Thanks Joe.
Saved me the writing.
I will take a slight overvoltage over droop any day of the week.
Also, if you're running that fine a line that you're worried about CPU destruction because of slight overvoltage from LLC, then you are at the end of your chain anyway. Time to pony up for some sub ambient cooling, or back your overclock down a little.
2 decades of hands on experience, you learn a little something.

 

[Zerileous]

Good to know guys. I was under the impression (IIRC due to a video by buildzoid) that LLC was there to protect the CPU from voltage spikes that software monitors are not sensitive enough to record (when the VRM is placed under a load, the voltage oscillates around the set level until it reaches a stable current). But if I think about it hypothetically, how is 0.1v of droop going to protect anything from a spike of 0.5v. In other words, if my specified or personal "safe" voltage for 24h operation is 1.35v and I achieve the maximum OC at that set voltage, and then by eliminating vdroop cause the CPU to see momentary spikes of 1.4v, it isn't going to hurt anything (because the actual short term ambient max of the CPU is 1.45v). And furthermore if the spike is actually large enough to do damage, then the vdroop is not actually enough to protect anything.

Am I on the right track here?

 

[wingman99]

Good to know guys. I was under the impression (IIRC due to a video by buildzoid) that LLC was there to protect the CPU from voltage spikes that software monitors are not sensitive enough to record (when the VRM is placed under a load, the voltage oscillates around the set level until it reaches a stable current). But if I think about it hypothetically, how is 0.1v of droop going to protect anything from a spike of 0.5v. In other words, if my specified or personal "safe" voltage for 24h operation is 1.35v and I achieve the maximum OC at that set voltage, and then by eliminating vdroop cause the CPU to see momentary spikes of 1.4v, it isn't going to hurt anything (because the actual short term ambient max of the CPU is 1.45v). And furthermore if the spike is actually large enough to do damage, then the vdroop is not actually enough to protect anything.

Am I on the right track here?

I use to try and believe those articles on the web that don't make sense that LLC is to protect the processor from voltage spikes. The reason it's not true because the PC gets the same voltage spikes with vdroop. What I have found over the years also recent research about load line here https://en.wikipedia.org/wiki/Load_line_(electronics) is the processor needs higher voltage with low load and lower voltage for higher load that AUTO LLC provides with vdroop.

A lot of folks find the same situation when testing and set the load line so there is no vdroop then set the same load core voltage that was on AUTO LLC and then just like you they have low load crashes or boot crashes. Because transistors need higher core voltage for low load do to the low current at low load compared high load.

So with LLC AUT0 then Vdroop I can have a little lower core voltage at high load.

 

[Xifios]

So should i then put my LLC to auto again but in turn also lower my voltage? Currently I have set it to 1.31V with LLC on medium which results in the voltages ranging from 1.27V to 1.308V.

Also if i were to leave it like that is this current configuration degrading my cpu significantly faster?
I still want to get at least a few more years out of this cpu before it dies and considering I have had this one since 2012 it already is somewhat old I guess.

 

[wingman99]

So should i then put my LLC to auto again but in turn also lower my voltage? Currently I have set it to 1.31V with LLC on medium which results in the voltages ranging from 1.27V to 1.308V.

Also if i were to leave it like that is this current configuration degrading my cpu significantly faster?
I still want to get at least a few more years out of this cpu before it dies and considering I have had this one since 2012 it already is somewhat old I guess.

It's a minor amount of voltage difference with playing around with LLC. So whatever works the best for you. At all your low voltage settings I have seen it will last for many more years. Folks are running 1.4v for many years.

 

[Xifios]

Ok thanks for the reasurrance I still need to wrap my head around all of this beeing a total noob when it comes to overclocking^^

Even though you said at these voltages it would still last me many more years I wonder if I should go back from 4.4 to 4.3ghz because @4.3 I had it stable with only 1.26V max.
Would you just leave it at 4.4 or is the performance increase so little that in regards of lifespan it is a lot more beneficial to go back to 4.3ghz @1.26V?

 

[Zerileous]

Truly only you can decide how much risk you can tolerate. We can tell you what it generally safe for your CPU, but even a stock CPU could give out tomorrow.

 

[Mr.Scott]

If it makes you feel better, 100 MHz, you will never notice in day to day usage. Only when benchmarking.