• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Disadvantage of LLC?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I quoted the author directly too except in my quote he says in plain English it may not be fast enough. You're just quoting numbers and applying your own guess work by claiming its more than enough.

I'm done owning you, time to watch movies. :sneaky:
 
It's not that the concept of LLC is flawed, it's the board manufacturers 1) messing it up, or 2) trying to deceit enthusiasts.

You may be able to find a sweet spot, but in order to achieve that physical measurement is a must. LLC can also affect any/every stage of power delivery to the CPU, and in that case you may even need to measure more than one spot on the board to accurately measure the actual vCore that CPU receives.

Most here will probably rely on software readings. They are inaccurate and cannot capture the degree of fluctuation quickly enough. It is possible, though, to get close the ball-park through patient trial-and-error's.
 
It's not that the concept of LLC is flawed, it's the board manufacturers 1) messing it up, or 2) trying to deceit enthusiasts.

You may be able to find a sweet spot, but in order to achieve that physical measurement is a must. LLC can also affect any/every stage of power delivery to the CPU, and in that case you may even need to measure more than one spot on the board to accurately measure the actual vCore that CPU receives.

Most here will probably rely on software readings. They are inaccurate and cannot capture the degree of fluctuation quickly enough. It is possible, though, to get close the ball-park through patient trial-and-error's.

Yea ok, and how do you propose we measure different spots on the board? Guess and Check? HAHAHAHAHA 😀😀
 
An update - I've found that although I need to increase my offset with LLC disabled, my lowest stable 100% load voltage as reported by CPU-Z is much lower. With LLC, the lowest I could get stable at 4.5ghz was approximately 1.280v. With LLC off, I've been stress-testing all day with a reported voltage of 1.248v. My uneducated hypothesis is that the times when the CPU needs the extra volts the most is when it's changing load and/or frequency, and during these times the voltage is actually higher with LLC disabled, which lets me run a lower voltage overall.
 
LLC is safe. PERIOD

It's amazing how you insist on spreading BS.

There's a reason why the lights dim momentarily in a room when you turn on a high current load on the same circuit, and then brighten momentarily when you turn the load off.

This is the same thing. It doesn't matter how well built the VRM is, a high current draw will always cause a voltage drop, and returning to idle will always cause a voltage spike.

If you're able to build a circuit that doesn't exhibit this behavior, then you can claim LLC is safe.
 
It's amazing how you insist on spreading BS.

There's a reason why the lights dim momentarily in a room when you turn on a high current load on the same circuit, and then brighten momentarily when you turn the load off.

This is the same thing. It doesn't matter how well built the VRM is, a high current draw will always cause a voltage drop, and returning to idle will always cause a voltage spike.

If you're able to build a circuit that doesn't exhibit this behavior, then you can claim LLC is safe.

For CPUs , voltage peaks are even a worse matter than for lights
wich can be seen as stable damped systems along with their current
source that is the main voltage.

The CPU supply use a feedback loop that not only has finite response time but also second order transfer function , meaning that when the system
try to correct voltage variation there will be overshoots in respect to the
final value , hence the faster the circuit the bigger the overshoots that can ultimately , if of too high amplitude , go beyond the CPU safe voltage.

Since this is an instantaneous variation , it will not be measured
as such by CPUZ or Everest , only the integral is measured.

It is well documented in EE engeenering textbooks.
 
Last edited:
No CPUz numbers are scaled. I can't seem to find that equation to "descale" it.

But it's around the forums somewhere.

LLC is safe. PERIOD

I suspect you are thinking of one of my posts about CPUz quantizing the Vcc reading in 0.008V increments.

Core Temp reports VID, CPUz reports Vcc.

And even then, CPUz doesn't report Vcc exactly, rather it reports something that is rounded to the nearest 0.008V increment.

In Excel the formula is as follows:
Code:
=ROUNDDOWN(([COLOR=darkgreen]A1[/COLOR]/0.001)/8,0)*8/1000
^ Where "A1" is the cell containing the actual Vcc per the BIOS (or multimeter if you have one setup).
Welcome to the forums as a non-lurker PowerK :thumbsup:

The answer to your first question is that it does not matter how the volts get to the CPU, provided it really is 1.200V, and provided the transients aren't the problem. (CPUZ is going to capture/report the transients)

Another thing to realize about CPUZ and voltage is that is does NOT report the CPU voltage, it reports something that is close to it but is rounded down by a value that can be as large as 0.008V.

Case 1 above could actually be 1.207 being rounded down to 1.200 in CPUZ while Case 2 could actually be 1.200 volts that isn't being rounded at all.

The forumula CPUz uses is the following: (forumla is in Excel notation, if you want to play with it)

CPUz Voltage = ROUNDDOWN((BIOS_Vcc/0.001)/8,0)*8/1000
 
An update - I've found that although I need to increase my offset with LLC disabled, my lowest stable 100% load voltage as reported by CPU-Z is much lower. With LLC, the lowest I could get stable at 4.5ghz was approximately 1.280v. With LLC off, I've been stress-testing all day with a reported voltage of 1.248v. My uneducated hypothesis is that the times when the CPU needs the extra volts the most is when it's changing load and/or frequency, and during these times the voltage is actually higher with LLC disabled, which lets me run a lower voltage overall.

I noticed the same thing but I wonder how accurate cpu-z is reading my core.
 
An update - I've found that although I need to increase my offset with LLC disabled, my lowest stable 100% load voltage as reported by CPU-Z is much lower. With LLC, the lowest I could get stable at 4.5ghz was approximately 1.280v. With LLC off, I've been stress-testing all day with a reported voltage of 1.248v. My uneducated hypothesis is that the times when the CPU needs the extra volts the most is when it's changing load and/or frequency, and during these times the voltage is actually higher with LLC disabled, which lets me run a lower voltage overall.

That's the exact opposite of what LLC does so I'm not so sure your readings are accurate. LLC reduces vdroop at load so your load voltage will be higher.
 
Which is exactly what he said as far as I can tell.....

As far as i can tell he said voltage at load is higher with LLC disabled which, again, is the opposite of what would happen with LLC disabled. LLC disabled = more vdroop. Vdroop = lower load voltage.
 
As far as i can tell he said voltage at load is higher with LLC disabled which, again, is the opposite of what would happen with LLC disabled. LLC disabled = more vdroop. Vdroop = lower load voltage.

I had to compensate for vdroop when turning LLC off (by raising the voltage) but the lowest stable voltage (at load) is lower with LLC off.
 
I had to compensate for vdroop when turning LLC off (by raising the voltage) but the lowest stable voltage (at load) is lower with LLC off.

Got it... Yeah that makes more sense than what i thought you said. One of the things you'll notice with your current settings is that when your CPU is running at full speed but not necessarily full load, your voltage is a lot higher than it needs to be since vdroop has not set in.
 
With LLC, the lowest I could get stable at 4.5ghz was approximately 1.280v. With LLC off, I've been stress-testing all day with a reported voltage of 1.248v.

As far as i can tell he said voltage at load is higher with LLC disabled which, again, is the opposite of what would happen with LLC disabled. LLC disabled = more vdroop. Vdroop = lower load voltage.

Last time I checked 1.280v is higher than 1.248v maybe things have changed though, damn laws of physics.
 
Last time I checked 1.280v is higher than 1.248v maybe things have changed though, damn laws of physics.

I was responding to this:

My uneducated hypothesis is that the times when the CPU needs the extra volts the most is when it's changing load and/or frequency, and during these times the voltage is actually higher with LLC disabled, which lets me run a lower voltage overall.

After his last explanation, I see that he was referring to the voltage he set in bios is higher with LLC disabled.

But if you want to be a smartass, try not to confuse physics with math. They aren't interchangeable.
 
Last edited:
All this argument over load line calibration is freaking pointless.

KEY ELEMENT, Idle voltage🙄

Your computer idles 80-90% of the time.

On an overclock: 4.8ghz

Without LLC, idle voltage will be 1.35v to 1.38v

With LLC, idle voltage will be 1.0v to 1.06v


There's nothing you can freaking do about this. 🙁🙁


LLC MUST BE ENABLED. 😎
 
Wow, somebody needs to learn to overclock properly if they are idling at 1.35v without LLC!

You can turn on C1e, but that doesn't kick in right away. :'(

And I'm not saying those are the differences literally, i'm just emphasizing it's by a huge margin.

Id you idle at 4.8ghz, you're going to be hitting 1.40v on low loads without LLC, and when it goes to large loads, the voltage drops to 1.33 or 1.35
 
Last edited:
Here's the exact situation

I need 1.33v minimum to be stable at 4.8ghz

Prime small ftt, not loading the memory controller,

Without LLC 1.31-1.32

With LLC 1.328-1.336

If it goes below 1.328 It's going to crash

If I engage the memory controllers in any application at this setting with LLC. I'll hit

With LLC 1.344v-1.35v

Without LLC, I'm hitting 1.36v+ because I have to use + offset to compensate for the "no memory controller-load" VID which is lower than my stable.


LLC works to give you LOWER voltage overall not HIGHER.
 
Lower through the spectrum below full load, correct.

I'm not sure what you mean by spectrum, but yea there is more than one full load, that's the problem.

Intel VID is what's screwing people over.

A better overclocking feature would be for the motherboard to determine voltage directly

So you can set a voltage for EVERY load condition and multiplier separately. :sneaky:

Yea, we'd have hundreds of value to test, but that's awesome. 😎
 
Back
Top