Proper 9900k LLC level? - Page 2 - Overclock.net - An Overclocking Community

Forum Jump: 

Proper 9900k LLC level?

Reply
 
Thread Tools
post #11 of 15 (permalink) Old 09-08-2019, 06:39 PM - Thread Starter
New to Overclock.net
 
Melodist's Avatar
 
Join Date: Sep 2016
Posts: 195
Rep: 1 (Unique: 1)
Quote: Originally Posted by Falkentyne View Post
If you're going to want to set a idle bios voltage of 1.4v, you're just better off switching to auto voltages, disabling loadline calibration (standard or LLC2, which I think is 1.6 mOhms on an Asus. LLC1 may be 2.1 mOhms which is too much vdroop on 8 core processors--intel specifies 1.6 max), and just using AC Loadline to control the idle and load voltages. You would get a lower idle voltage (and with same load voltage at 5 ghz) with an AC Loadline of 1.0 mOhms and LLC Standard/Normal, than you would with fixed 1.40v voltage with LLC Standard.
I'll try that, it might end up with a better stability and lower temps

But pushing 12k prime avx is nonesense, Aida64 has proved to me to be the most reliable AVX stresstest, just running FPU, which stresses more than CPU and FPU.
Melodist is offline  
Sponsored Links
Advertisement
 
post #12 of 15 (permalink) Old 09-09-2019, 06:26 AM - Thread Starter
New to Overclock.net
 
Melodist's Avatar
 
Join Date: Sep 2016
Posts: 195
Rep: 1 (Unique: 1)
Quote: Originally Posted by Falkentyne View Post
Quote: Originally Posted by Melodist View Post
But on a second thought, what keeps me then from going even lower LLC wise like medium and setting s voltage of somewhere 1.4 volts? Which would be even more stable and cool but where is the limit and where does it translate how?
If you're going to want to set a idle bios voltage of 1.4v, you're just better off switching to auto voltages, disabling loadline calibration (standard or LLC2, which I think is 1.6 mOhms on an Asus. LLC1 may be 2.1 mOhms which is too much vdroop on 8 core processors--intel specifies 1.6 max), and just using AC Loadline to control the idle and load voltages. You would get a lower idle voltage (and with same load voltage at 5 ghz) with an AC Loadline of 1.0 mOhms and LLC Standard/Normal, than you would with fixed 1.40v voltage with LLC Standard.
Is there a tutorial for a baseline?
Melodist is offline  
post #13 of 15 (permalink) Old 09-09-2019, 09:52 AM - Thread Starter
New to Overclock.net
 
Melodist's Avatar
 
Join Date: Sep 2016
Posts: 195
Rep: 1 (Unique: 1)
This a great tutorial:

https://forums.bit-tech.net/index.ph...master.353729/

Not sure about the power saving ac line calibration though.
Melodist is offline  
Sponsored Links
Advertisement
 
post #14 of 15 (permalink) Old 09-09-2019, 12:42 PM
Facepalm
 
Falkentyne's Avatar
 
Join Date: Dec 2007
Location: Riverside
Posts: 7,523
Rep: 447 (Unique: 311)
Quote: Originally Posted by Melodist View Post
Is there a tutorial for a baseline?
No, because no one except me has even considered these things before. No one ever actually looked at the Intel document sheets and tried to understand the relationship between max VID and max CURRENT.
It became sort of clear when in the electrical specifications area, Intel lists Max DC_LL (DC Loadline), as 1.6 mOhms and calls it "Loadline slope within the VR Loop capability", meaning this specifically refers to "VRM Loadline". It also doesn't give any minimum, only a maximum. Then under AC_LL, it just says "AC Loadline" and "Same as max DC_LL". The notes below mention however:

13. LL specification values should not be exceeded. If exceeded, power, performance and reliability penalty are expected.
14. Load Line (AC/DC) should be measured by the VRTT tool and programmed accordingly via the BIOS Load Line override setup
options. AC/DC Load Line BIOS programming directly affects operating voltages (AC) and power measurements (DC). A
superior board design with a shallower AC Load Line can improve on power, performance, and thermals compared to boards
designed for POR impedance

The problem is this gets confusing because they say in the note that "DC Loadline" affects power measurements while AC Loadline affects operating voltages. But in the DC_LL chart, it says "loadline slope within the VR Loop", which is clearly NOT the same thing! So it sounds like the "loadline slope" is the default VRM Loadline that is designed to be used for 8 core processors (1.6 mOhms), and DC Loadline has to be set to the same value as VRM Loadline (in mOhms) so that the CPU Package Power remains accurate (VID * Amps), since lowering DC Loadline will cause the VID to droop less (the same way increasing loadline calibration --VRM Loadline--makes the VCCcore drop less).

I got more bored last night and, with auto voltages, (and AC Loadline set to 1.0 mOhms, whatever), I decided to disable 2 cores while Loadline calibration was set on Standard.

Checking the VR VOUT and VID afterwards, I noticed that now VR VOUT was lower than VID again. I had DC Loadline set to 1.6 mOhms. When I set it to 2.1 mOhms, VID and VR VOUT were exactly the same! And in the intel document sheet, DC_LL is 2.1 mOhms for 4 and 6 core processors! So...there we go! LLC=Standard sets the VRM Loadline slope to 1.6 mOhms on 8 cores and 2.1 mOhms on 4 and 6 cores! Note that on 6 cores, max amps is now 138 amps on 95W SKU's (instead of 193) and for 4 cores, it's 100 amps (91 watt SKU's).

It's very easy then to determine the absolute maximum safe voltage (up to 99C) when you know all of these values (VRM Loadline slope mOhms, max VID in millivolts and max amps).

As I said above you can start like this:

Vhigh=Bios target voltage set on manual voltage, or VRM target voltage @ full load on auto voltage (set DC Loadline=1 and look at VID to find this on auto voltage)

vlow=minimum load voltage at most amps (on-die sense or VR VOUT measured only)

Vhigh-vlow=max vdroop (millivolts)

R=loadline slope resistance (mOhms)
(Vhigh-Vlow) / R = Amps
Vdroop / R = Amps.
Vdroop / Amps = R

Then you can calculate whatever your maximum safe voltage should be at a given amps value (if amps is known)
by using the formula (1520mv - ( VRM Loadline slope * Amps))

If you don't know your amps, you can calculate it by vdroop divided by VRM loadline resistance (provided your voltage measurements are accurate).

The nice thing about the above formula is, you can plug in your custom loadline calibration resistance values instead of the 1.6 or 2.1 mOhms Intel default loadlines, and your VRM target voltage, to tell you what the vdroop should be at full load, and then compare that to all of the absolute max baselines at "default" loadline calibration. Then you'll see if your loadline is too aggressive or your bios voltage is too high.

like:
193 amps: 1520 - (1.6 * 193) = 1.213v
150 amps: 1520 - (1.6 * 150) = 1.280v
125 amps: 1520 - (1.6 * 125) = 1.320v
100 amps: 1520 - (1.6 * 100) = 1.360v

etc

Now let's say you're using a fixed bios voltage and loadline calibration High/LLC6 (0.4 mOhms), and 1.35v:
You're pulling 100 amps:

1350 - (0.4 * 100) = 1310mv load voltage. Looking at the above chart for 100 amps, you're good.

150 amps:
1350 - (0.4 * 150) = 1.290v.

Time to be careful huh?

Also remember that the more LLC you use, the larger the transient spikes are and the larger the transient dips are. Dips could crash you at heavier amps if your RMS voltage is too slow to borderline, spikes could cause slow degradation.

https://www.overclock.net/forum/6-in...l#post28022572

[email protected] ghz, RX Vega 64, 32GB DDR4, Gigabyte Aorus Master, Seasonic Platinum 1000W, Corsair 760T
Alt: MSI GT73VR Throttlebook with 7820HK @ 4.7 ghz, GTX 1070 MXM TDP mod to 230W, 32 GB RAM

Quote: Originally Posted by sakete View Post
Well, I want you to know I have an academic degree in speculation.
Falkentyne is offline  
post #15 of 15 (permalink) Old 05-15-2020, 06:24 AM
New to Overclock.net
 
Join Date: Sep 2019
Posts: 3
Rep: 0
Quote: Originally Posted by Falkentyne View Post
1.335v at idle is nothing because the amps draw at 1.335v is extremely low at idle (with LLC5/LLC High that's probably actually 1.320v-1.325v at idle as there is vdroop even at idle). Fully safe (even safer with lower LLC's!). Now if you were pulling 193 amps at 1.335v load (Measured via CPU on-die differential sense), you will degrade that processor (plus it would be completely uncoolable as well).

While 1.7v idle will slowly burn up your processor for sure, your processor is safe as long as you respect Intel's amps/voltage curve. (the problem is Loadline Calibration skews the entire formula because Intel doesn't design for Loadline calibration, only for AC Loadline (max 1.6 Ohms).

I don't believe that VRM Loadline is even mentioned in the Intel specification documents, only AC Loadline (CPU power supply) and DC Loadline (Power measurements). VRM Loadline is what is set by Loadline Calibration. It mentions that a lower AC Loadline will improve on current and temps (this will reduce the load voltage).

To stay in design limits, VRM input voltage must not exceed 1.520v, and VRM current (Amps) must not exceed 193 amps on 8 core CFL, and AC Loadline set to maximum of 1.6 mOhms. All of this assumes NO loadline calibration is being used. The 1.520v value is important because the VRM actually receives this input voltage on higher processor speeds/AC Loadlines when using AUTO voltage. Vdroop is what saves you.

The formula is basically this, with resistance being your loadline value (VRM loadline). While AC and DC loadlines are also in resistances, they are not used in the formula.

1520(mv) - (Amps * resistance)=maximum safe voltage your CPU should not exceed.
so at 193 amps, that's 1520 - (193 * 1.6)= 1.213v
At 150 amps that's 1520 - (150 * 1.6)= 1.280v
Thank you very much for your explanation. Finally, this LLC issue is becoming clearer to me!

Just to see if I am in the right understanding... currently my configuration is as follows:

9900ks
core: 51x
uncore: 48x
LLC 4
Adaptive 1.360v
(Asus Maximus XI Formula - LLC 1-8)

My load voltage is 1.181v on Prime95 (FFTs 12k and without AVX), and pulls 133A.

Therefore, according to the formula, the maximum safe voltage would be:

1.520 - (133 * 1.6) = 1.307v

Please, now my question is:
Would this be the maximum safe voltage at full load or the maximum configured in the BIOS?
Should I raise the LLC to 5 and lower the BIOS voltage to a maximum of 1,307 !? Or am I fine at 1.360v @ LLC 4?!

Thank you for your help!
ncpneto is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off