9900k: large Vdroop. Is load voltage the main 1 to watch? - Page 2 - Overclock.net - An Overclocking Community
Forum Jump: 

9900k: large Vdroop. Is load voltage the main 1 to watch?

Reply
 
Thread Tools
post #11 of 20 (permalink) Old 11-28-2018, 09:42 PM - Thread Starter
New to Overclock.net
 
Join Date: Feb 2018
Posts: 28
Rep: 0
Quote: Originally Posted by Falkentyne View Post
Quote: Originally Posted by Jaz11 View Post
Hi. My maximus xi formula has large vdroop. LLC6 1.37v bios down to 1.26v under load. Is voltage under load the main 1 to watch? Is it fine to start settings 1.41v in bios to push for 5.1ghz seeing as vdroop will end up at around 1.29-1.3v?

Llc7 improves the vdroop by about .03 but bring extra heat.

Any help qould be great thanks.
What is the highest LLC on your mainboard?
Are you using static or adaptive (normal / auto) voltages?
1.37v idle to 1.26v load is massive. Have you tested the voltages with a digital multimeter to see if the sensors are wrong?
In your "1.37v idle, 1.26v load", does your "CPU VID" also read 1.37v idle and 1.26v load (VID, not vcore?).

1.37v to 1.26v is basically NO LLC at all.
What happens if you use LLC5 or LLC4?
Test the others. Perhaps LLC6 is just busted and not working?
Llc8 is the maximum but it also doesnt change voltage, crashed straight away for some reason.

1.37v is the bios voltage, idle goes around 1.35v then the 1.26v. This is all at LLC6 and manual voltage. All cpu currents maxed out from der8aurs 9900k guide
Jaz11 is offline  
Sponsored Links
Advertisement
 
post #12 of 20 (permalink) Old 11-28-2018, 09:45 PM - Thread Starter
New to Overclock.net
 
Join Date: Feb 2018
Posts: 28
Rep: 0
All cpu powers and currents are maxed out as per der8aur 9900k guide. Latest bios, 0506 from memory it was
Jaz11 is offline  
post #13 of 20 (permalink) Old 11-28-2018, 09:49 PM
Facepalm
 
Falkentyne's Avatar
 
Join Date: Dec 2007
Location: Riverside
Posts: 7,369
Rep: 441 (Unique: 306)
Quote: Originally Posted by elmor View Post
The LLC settings and resulting droop on M11 are the same as on M10, but with better SIO readings. From what I've seen other boards are showing values similar to M10.

SIO (traditional) in the graph refers to Maximus X Hero.



For example: If you set Manual Mode Voltage = 1.300V + Load-line Calibration = Level 6 on Maximus X Hero, CPU-Z or other software will read 1.312V. On Maximus XI Hero the reading will be 1.190V. The actual voltage the CPU is getting is 1.205V in both cases. The difference is amplified at larger currents, with 9900K the difference is larger than with previous CPUs.
This is probably the most interesting news I've read all year. I am not exaggerating.
So that means that that "+100mv" rumor that was going around was from the traditional SIO reading?
Which also means that that one person with that stable 9900K at 1.21v at 5 ghz was actually getting around that voltage then....

But how do the internal IA AC DC loadlines come into the picture then? (the 1.60 mOhms reference value for 8 core CPU's)?

When using pure adaptive voltage, (onboard loadline calibration set to disabled / normal), the cpu vcore is set based on the default VID. And the socket MLCC reading, which I assume is what the Gigabyte Sensor #2 is reading, is extremely close to what the cpu's VID is showing.

As in this test I did here:

So what is the true cpu voltage then? And in that case that the CPU were getting much lower voltage than the MLCC reading, why isn't the VID much lower in that case?
Attached Thumbnails
Click image for larger version

Name:	4700mhz_load_noAVX_adaptive.jpg
Views:	226
Size:	713.4 KB
ID:	234814  


[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 online now  
Sponsored Links
Advertisement
 
post #14 of 20 (permalink) Old 11-28-2018, 10:43 PM
New to Overclock.net
 
elmor's Avatar
 
Join Date: Mar 2014
Posts: 873
Rep: 354 (Unique: 158)
Quote: Originally Posted by Falkentyne View Post
This is probably the most interesting news I've read all year. I am not exaggerating.
So that means that that "+100mv" rumor that was going around was from the traditional SIO reading?
Which also means that that one person with that stable 9900K at 1.21v at 5 ghz was actually getting around that voltage then....

But how do the internal IA AC DC loadlines come into the picture then? (the 1.60 mOhms reference value for 8 core CPU's)?

When using pure adaptive voltage, (onboard loadline calibration set to disabled / normal), the cpu vcore is set based on the default VID. And the socket MLCC reading, which I assume is what the Gigabyte Sensor #2 is reading, is extremely close to what the cpu's VID is showing.

As in this test I did here:

So what is the true cpu voltage then? And in that case that the CPU were getting much lower voltage than the MLCC reading, why isn't the VID much lower in that case?
I'm not sure which "+100mv" rumor you're talking about.

Quote: Originally Posted by Silicon Lottery View Post
Another note, the Asus XI Hero is producing the same overclocks at a "lower voltage" compared to other Z390 boards, but with higher load temperatures. I'm assuming it is not very accurate in reporting what voltage is actually being applied.
This quote shows the result of the improved readings, where it looks like lower voltage on M11 but it's really other boards that read too high.

IA AC/DC loadline are load-line values that are applied to Intel SVID requests, and thus only affects Adaptive/Offset voltage modes. They are tuned per board, the 1.6mOhm recommendation from Intel means that the board design should result in a value lower than this.

If a CPU runs at a certain frequency with a certain load, it has a table specifying the required voltage for that scenario (VID). Let's say the CPU frequency is 4.7 GHz, the output current 100A and the required VID is 1.300V. The IA DC load-line value has been set to 1.0 mOhm. Using this IA DC load-line value, the CPU will compensate for board losses and VRM load-line by applying an offset to the requested VID. The theoretical voltage drop will be load-line*output = 1.0 mOhm * 100A = 100mV. Probably then the CPU will request something like 1.300+0.1 = 1.400V. Then there's also the Load-line Calibration BIOS option which changes the VRM load-line value and probably also the IA AC/DC load-line Auto values. The different "levels" are fixed sets of tuned load-line mOhm values. AC load-line values are for transients and add even further complexity.

The VID value is what the CPU requests, not what it gets. Your actual CPU Voltage will be slightly lower than the MLCC reading. Does HWInfo pick up the VRM controller (IR35201) as a separate sensor on your board? That should provide a very accurate voltage.

Last edited by elmor; 11-28-2018 at 10:50 PM.
elmor is offline  
post #15 of 20 (permalink) Old 11-28-2018, 10:54 PM
Facepalm
 
Falkentyne's Avatar
 
Join Date: Dec 2007
Location: Riverside
Posts: 7,369
Rep: 441 (Unique: 306)
Quote: Originally Posted by elmor View Post
I'm not sure which "+100mv" rumor you're talking about.



This quote shows the result of the improved readings, where it looks like lower voltage on M11 but it's really other boards that read too high.

IA AC/DC loadline are load-line values that are applied to Intel SVID requests, and thus only affects Adaptive/Offset voltage modes. They are tuned per board, the 1.6mOhm recommendation from Intel means that the board design should result in a value lower than this.

If a CPU runs at a certain frequency with a certain load, it has a table specifying the required voltage for that scenario (VID). Let's say the CPU frequency is 4.7 GHz, the output current 100A and the required VID is 1.300V. The IA DC load-line value has been set to 1.0 mOhm. Using this IA DC load-line value, the CPU will compensate for board losses and VRM load-line by applying an offset to the requested VID. The theoretical voltage drop will be load-line*output = 1.0 mOhm * 100A = 100mV. Probably then the CPU will request something like 1.300+0.1 = 1.400V. Then there's also the Load-line Calibration BIOS option which changes the VRM load-line value and probably also the IA AC/DC load-line Auto values. The different "levels" are fixed sets of tuned load-line mOhm values. AC load-line values are for transients and add even further complexity.

The VID value is what the CPU requests, not what it gets. Your actual CPU Voltage will be slightly lower than the MLCC reading. Does HWInfo pick up the VRM controller (IR35201) as a separate sensor on your board? That should provide a very accurate voltage.
Learned a lot today from you, Elmor. Probably more than i've learned all year. Thank you. I was always wondering how that formula worked.
(At least I always tell people NEVER to go near the higher values for the IA AC DC loadlines).

Anyway this 100mv rumor has been spread all over. And your chart basically puts it all to rest completely with proof.
Here's the threads.

https://rog.asus.com/forum/showthrea...n-100mv-offset
https://www.reddit.com/r/intel/comme..._100mv_offset/

There was also a thread on a German or i dont know some strange european language with de (?) site and a chinese site too.

There was something brought up here about that too, but referencing those.

Yes HWInfo picks up IR35201 as a separate sensor.
The idle vcore voltage shown on the more accurate vcore sensor--there are two, the (ITE IT8792E) is more accurate than IT 8688E), is equal to the "VR Out" (IR 35201) voltage at idle.
At full load (I am using LLC=turbo), the ITE IT8792E sensor is about the same as idle, while VR Out drops a whole lot.

Right now when writing this, I have vcore set to 1.270 in the bios.
Idle shows up as 1.265v for the ITE vcore sensor and "VR VOUT".
Full prime95 load (No AVX) shows 1.276v for the ITE ITE8792E Vcore sensor and 1.230v for VR Vout (IR 35201).

Are you saying 1.230v is the true voltage the CPU is getting?

I thought VR Vout was the vcore that "would" be the voltage if I had loadline calibration disabled...

[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.

Last edited by Falkentyne; 11-28-2018 at 11:16 PM.
Falkentyne is online now  
post #16 of 20 (permalink) Old 11-28-2018, 11:20 PM
New to Overclock.net
 
Raghar's Avatar
 
Join Date: Jun 2012
Posts: 1,901
Rep: 38 (Unique: 32)
Quote: Originally Posted by elmor View Post
I'm not sure which "+100mv" rumor you're talking about.
Few years ago one owner of Haswell CPU tested his CPU on manual and adaptive voltage, and was very livid when he found adaptive voltage adds 0.1 V when he uses program with SSE2 instructions.

Nowadays he's working as tech support/PR for one of whole PC sellers. And he's VERY adamant in using non-K CPU and avoiding any experimentation. It's scary how some people when grow up and gets a job starts to be risk avoidant.

(He was also one of these boys who discovered that certain Asus board was warping when he placed heavier air cooler on it. He got it exchanged to different Asus board type which didn't warped...)

Frankly SSE2 was new on Haswell, and needed quite a bit of juice. Thus CPU asking for more juice when it uses SSE2 is logical. The question is if current CPUs are doing this as well, no FIVR means they would need ask MB.
Raghar is offline  
post #17 of 20 (permalink) Old 11-28-2018, 11:23 PM
New to Overclock.net
 
elmor's Avatar
 
Join Date: Mar 2014
Posts: 873
Rep: 354 (Unique: 158)
Quote: Originally Posted by Falkentyne View Post
Learned a lot today from you, Elmor. Probably more than i've learned all year. Thank you. I was always wondering how that formula worked.
(At least I always tell people NEVER to go near the higher values for the IA AC DC loadlines).

Anyway this 100mv rumor has been spread all over. And your chart basically puts it all to rest completely with proof.
Here's the threads.

https://rog.asus.com/forum/showthrea...n-100mv-offset
https://www.reddit.com/r/intel/comme..._100mv_offset/

There was also a thread on a German or i dont know some strange european language with de (?) site and a chinese site too.

There was something brought up here about that too, but referencing those.

Yes HWInfo picks up IR35201 as a separate sensor.
The idle vcore voltage shown on the more accurate vcore sensor--there are two, the (ITE IT8792E) is more accurate than IT 8688E), is equal to the "VR Out" (IR 35201) voltage at idle.
At full load (I am using LLC=turbo), the ITE IT8792E sensor is about the same as idle, while VR Out drops a whole lot.

Right now when writing this, I have vcore set to 1.270 in the bios.
Idle shows up as 1.265v for the ITE vcore sensor and "VR VOUT".
Full prime95 load (No AVX) shows 1.276v for the ITE ITE8792E Vcore sensor and 1.230v for VR Vout (IR 35201).

Are you saying 1.230v is the true voltage the CPU is getting?

I thought VR Vout was the vcore that "would" be the voltage if I had loadline calibration disabled...

Well, that's the same guy posting on two different sites. And yes, there's no "offset" just difference in readings. Same actual voltage in both cases.

The IR35201 VR Vout will be very accurate, you should rely on that value for your testing. It's the VRM controller ADC value, which is using CPU on-die sense. In this case, "LLC=Turbo" + 1.270V set in BIOS on your Gigabyte board seems similar to Load-line Calibration = Level 7 on M11.
elmor is offline  
post #18 of 20 (permalink) Old 05-09-2020, 06:36 PM
New to Overclock.net
 
Join Date: Jun 2019
Posts: 3
Rep: 0
Quote: Originally Posted by elmor View Post
Well, that's the same guy posting on two different sites. And yes, there's no "offset" just difference in readings. Same actual voltage in both cases.

The IR35201 VR Vout will be very accurate, you should rely on that value for your testing. It's the VRM controller ADC value, which is using CPU on-die sense. In this case, "LLC=Turbo" + 1.270V set in BIOS on your Gigabyte board seems similar to Load-line Calibration = Level 7 on M11.
Sorry for necroing this thread, but I have a question about this "True CPU Vcore" or as it is called in Hwinfo "VR Vout" Does this mean we should keep this value less than 1.45V on 8th gen core I5, or is it still the Bios manual Voltage we should keep under 1.45V. I ask, because that means I can in fact OC my 8600k to 5 ghz. I managed to do a stable 1 hour test with Real bench at that speed, but I didn't like the Vcore Readouts (I believe it was a brief max of 1.456V) I was getting so I went back down to 4.9 ghz. This all happened before I happened to stumble across this thread and now I am wondering if I acted too hastily in my decision. The Temperatures I was getting while gaming were a smidge over 85.
Rob Martin is offline  
post #19 of 20 (permalink) Old 05-09-2020, 07:31 PM
Facepalm
 
Falkentyne's Avatar
 
Join Date: Dec 2007
Location: Riverside
Posts: 7,369
Rep: 441 (Unique: 306)
Quote: Originally Posted by Rob Martin View Post
Sorry for necroing this thread, but I have a question about this "True CPU Vcore" or as it is called in Hwinfo "VR Vout" Does this mean we should keep this value less than 1.45V on 8th gen core I5, or is it still the Bios manual Voltage we should keep under 1.45V. I ask, because that means I can in fact OC my 8600k to 5 ghz. I managed to do a stable 1 hour test with Real bench at that speed, but I didn't like the Vcore Readouts (I believe it was a brief max of 1.456V) I was getting so I went back down to 4.9 ghz. This all happened before I happened to stumble across this thread and now I am wondering if I acted too hastily in my decision. The Temperatures I was getting while gaming were a smidge over 85.
Depends on current.
For 8600k, find your current and use the target voltage based on that current as:
1520mv - (2.1 * amps)=load voltage that you do NOT want to exceed.

So if you were pulling 100 amps, your load voltage should not exceed 1.31v.

But as a general idea, no matter how droopy your vdroop is (even if it's at intel defaults), you shouldn't be sitting around with an idle voltage of 1.45v all the time. That's a bit pushing it. With a healthy loadline, I wouldn't like to go past 1.42v idle. And if you remove vdroop, your idle should be even lower (depending on how much vdroop you are reducing).

[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.

Last edited by Falkentyne; 05-09-2020 at 07:37 PM.
Falkentyne is online now  
post #20 of 20 (permalink) Old 05-14-2020, 01:24 AM
New to Overclock.net
 
Join Date: Jun 2019
Posts: 3
Rep: 0
Quote: Originally Posted by Falkentyne View Post
Depends on current.
For 8600k, find your current and use the target voltage based on that current as:
1520mv - (2.1 * amps)=load voltage that you do NOT want to exceed.

So if you were pulling 100 amps, your load voltage should not exceed 1.31v.

But as a general idea, no matter how droopy your vdroop is (even if it's at intel defaults), you shouldn't be sitting around with an idle voltage of 1.45v all the time. That's a bit pushing it. With a healthy loadline, I wouldn't like to go past 1.42v idle. And if you remove vdroop, your idle should be even lower (depending on how much vdroop you are reducing).
Yeah thanks for your reply, if my math is correct it looks like 4.9ghz is the best I will get with this chip. I can't really justify the risk for a measly extra 100mhz so I will keep an eye on it because the warm months are coming and I don't have air conditioning here.

Thanks for your help

"May our temperatures be low and our frame rates high"
Rob Martin 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