Z370 / Z390 VRM Discussion Thread - Page 395 - Overclock.net - An Overclocking Community

Forum Jump: 

Z370 / Z390 VRM Discussion Thread

Reply
 
Thread Tools
post #3941 of 4007 (permalink) Old 02-12-2019, 03:06 PM
New to Overclock.net
 
wingman99's Avatar
 
Join Date: Apr 2012
Posts: 2,999
Rep: 49 (Unique: 40)
Quote: Originally Posted by Falkentyne View Post
CPU package power is based on CPU VID remember?
No CPU package power is based on voltage X AMPs = watts measurement from the motherboard sensors. 1.32v X 130amp=171.6 watts

VID is (voltage identification digital) from the processor and is not used when using fixed core voltage only used with adaptive core voltage.

i5 8600K OC 5.0GHz
Noctua NH-U14S
Motherboard Gigabyte Z370-HD3
G.SKILL Ripjaws v 16GB F4-3200C14D-16GVK XMP Speed 3200 14-14-14-34
EVGA GeForce RTX 2070 XC ULTRA GAMING
wingman99 is online now  
Sponsored Links
Advertisement
 
post #3942 of 4007 (permalink) Old 02-12-2019, 03:16 PM
New to Overclock.net
 
Falkentyne's Avatar
 
Join Date: Dec 2007
Location: Riverside
Posts: 5,416
Rep: 362 (Unique: 258)
Quote: Originally Posted by wingman99 View Post
No CPU package power is based on voltage X AMPs = watts measurement from the motherboard sensors. 1.32v X 130amp=171.6 watts

VID is (voltage identification digital) from the processor and is not used when using fixed core voltage only used with adaptive core voltage.
You are half right and half wrong.

VID is still used for package power consumption.
It's not used for CPU Vcore when set to fixed, EXCEPT in most laptops, where "fixed" vcore acts like a VID override rather than a vcore override.
Unclewebb had the documentation for this. He also has more information I am not going to discuss here as I am not allowed to.
If it were not used, changing IMON SLOPE and IMON OFFSET would not spoof the CPU power reporting.

*Edit* Im aware I don't have the CPU speed shown in picture 1, I do not know how to have HWinfo multiple windows show the CPU VID CPU Package power AND VR output in 3 windows because you can only scroll the last window.

Test 1: IA AC Loadline=1, IA DC Loadline=1, 4.7 ghz core, x44 ring, VID 1.183v, Package power 198W, POUT (VRM) 181W, VR VOUT 1.131V, Amps 161.5W (IOUT).
(Notice that CPU package power is very close to 1.183 * 161.5?)

Test 2: IA AC Loadine=1, IA DC loadline=210, core x47, ring x36, VID 0.895v, CPU Package power 146W POUT (VRM) 178W, VR Vcore 1.133v, IOUT 158.750 amps.
(Notice that CPU package power is VERY VERY close to 158 * 0.895?)

There's your proof.
Attached Thumbnails
Click image for larger version

Name:	vidtest1.jpg
Views:	19
Size:	712.3 KB
ID:	253190  

Click image for larger version

Name:	vidtest2.jpg
Views:	17
Size:	736.8 KB
ID:	253192  


[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

Last edited by Falkentyne; 02-12-2019 at 05:27 PM.
Falkentyne is online now  
post #3943 of 4007 (permalink) Old 02-12-2019, 03:47 PM
New to Overclock.net
 
wingman99's Avatar
 
Join Date: Apr 2012
Posts: 2,999
Rep: 49 (Unique: 40)
Quote: Originally Posted by Falkentyne View Post
Spoiler!
Thanks for the information. I got close to the same example.

i5 8600K OC 5.0GHz
Noctua NH-U14S
Motherboard Gigabyte Z370-HD3
G.SKILL Ripjaws v 16GB F4-3200C14D-16GVK XMP Speed 3200 14-14-14-34
EVGA GeForce RTX 2070 XC ULTRA GAMING
wingman99 is online now  
Sponsored Links
Advertisement
 
post #3944 of 4007 (permalink) Old 02-12-2019, 05:11 PM
New to Overclock.net
 
Falkentyne's Avatar
 
Join Date: Dec 2007
Location: Riverside
Posts: 5,416
Rep: 362 (Unique: 258)
Quote: Originally Posted by wingman99 View Post
Thanks for the information. I got close to the same example.
It's fine.
But this is also playing havoc with Tibur Born (whatever his name is)'s results because he uses c-states, and c-states completely screw up the VRM reporting, because the CPU basically "goes to sleep" and ignores what the VRM's are trying to output. The only way to get reliable reporting from the VRM's when also using adaptive (or offset/speed shift/underclocking) is to keep c-states enabled and just use speed shift (Throttlestop 8.70 can do this easily for you).

But yeah, CPU VID is never that far away from CPU Voltage, unless you use a low cache speed when overclocking. CPU VID is linked to the cache voltage at certain mhz steps (starting from 800 mhz and going up to 4.7 ghz on 9900K's, although not every +100 mhz is a VID increase). That's why you have the ratio of cache wanting to be -3 lower than core speed. Going too high on the cache will raise the default VID (up to 4.7 ghz) but will hurt baseline stability if it's the same speed as the core. Setting the cache very low when using offset/auto/adaptive voltages will require a large voltage +offset or require NOT using IA AC loadline=1.

While I mostly know how DC loadline works (it works the same way, with respect to the CPU VID and current, as Loadline Calibration works with respect to the Vcore--higher levels of loadline calibration work like "reducing the mOhms level of the DC loadline value, so a 0 mOhm Loadline Calibration (Ultra Extreme, or LLC8) is similar to IA DC loadline=1 (0.01 mOhms).

I still don't know exactly how AC loadline works, but it directly affects the CPU power supply in some way (at idle, the VID is much higher with AC loadline=1.6 mOhms (160 in gigabyte bioses or MSI (maybe) Bioses), compared to AC loadline=0.01 mOhms (1 in Gigabyte bioses/MSI).

[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

Last edited by Falkentyne; 02-12-2019 at 05:21 PM.
Falkentyne is online now  
post #3945 of 4007 (permalink) Old 02-14-2019, 04:25 AM
New to Overclock.net
 
Timur Born's Avatar
 
Join Date: Oct 2015
Posts: 1,263
Rep: 107 (Unique: 49)
Quote: Originally Posted by Falkentyne View Post
CPU package power is based on CPU VID remember?
And no, it doesn't line up with the VRM controller's amp measurement because the CPU gets the amp load from somewhere else other than the VRM.

The only person who would know is either Elmor, or an Intel or possibly one of the Asus engineers. Very few on these forums would know.
Power (POUT) is based on VR VOUT (CPU on-die sense voltage) * Current (IOUT).

CPU package power is HEAVILY based on the CPU VID.
I can live with that. So the reported 250w CPU package power are higher than what the CPU really uses. POUT reports 220w instead, which means that the difference between real CPU power usage and power measured at the wall is even bigger (even after subtracting other components like PSU efficiency, fans, SSD and memory).

Which again begs the question: Is the remaining 100w turned into heat at the VRM stage?
Timur Born is offline  
post #3946 of 4007 (permalink) Old 02-14-2019, 04:30 AM
New to Overclock.net
 
Timur Born's Avatar
 
Join Date: Oct 2015
Posts: 1,263
Rep: 107 (Unique: 49)
Quote: Originally Posted by Falkentyne View Post
But this is also playing havoc with Tibur Born (whatever his name is)'s results because he uses c-states, and c-states completely screw up the VRM reporting, because the CPU basically "goes to sleep" and ignores what the VRM's are trying to output. The only way to get reliable reporting from the VRM's when also using adaptive (or offset/speed shift/underclocking) is to keep c-states enabled and just use speed shift (Throttlestop 8.70 can do this easily for you).
I am using adaptive voltage in combination with speed shift. VOUT/POUT reports mostly only get screwed when active C-states are combined with active Speedshift, by the way. Without Speedshift things are a bit different.

This doesn't matter for full load scenarios, of course, because at 100% load neither C-states nor Speedshift are invoked.
Timur Born is offline  
post #3947 of 4007 (permalink) Old 02-16-2019, 02:10 AM
New to Overclock.net
 
Timur Born's Avatar
 
Join Date: Oct 2015
Posts: 1,263
Rep: 107 (Unique: 49)
At 80°C the VRM on my Aorus Master draws up to 10 watts more power for the same workload compared to at around 40°C. I would have to test specifically at what temp it begins to worsen gradually.

Last edited by Timur Born; 02-16-2019 at 02:49 AM.
Timur Born is offline  
post #3948 of 4007 (permalink) Old 02-16-2019, 03:21 AM
New to Overclock.net
 
asdkj1740's Avatar
 
Join Date: Dec 2015
Posts: 956
Rep: 12 (Unique: 6)
https://www.chiphell.com/thread-1959959-1-1.html

toppc lin explains why msi has the 100.8 bclk settings instead of 100.
it is triggered when users have 2 sticks of ddr4 installed and xmp is enabled.
100.8 bclk helps 2 sticks ddr4 oc higher when xmp is enabled.
msi though lots of users wont tweak the small timings settings so msi set 100.8 bclk to help general users to oc higher without messing with those timings.
but now msi is going to disable this feature by default in the new coming bios.

"順便解釋一下 BCLK 100.8
這是為了突顯在開啟XMP情況下
一般市售的內存
在只插兩根時...可以超比較高
所以會自動微加0.8 BCLK(當時沒料到..大陸這邊.手動超的人 還真不少)
不喜歡的人可以手調100.6或99.98
GODLIKE 1.6x的BIOS我會加入選項.讓大家選擇是 跑100.8
(1.5版已經來不及加了)"

i remember toppc said nowadays tweaking bclk for cpu oc wont have any benefit over tweaking cpu ratio. so the bclk 100.8 is not for cpu oc / cheating with cpu scores.

Last edited by asdkj1740; 02-16-2019 at 03:27 AM.
asdkj1740 is offline  
post #3949 of 4007 (permalink) Old 02-17-2019, 04:02 AM
New to Overclock.net
 
Timur Born's Avatar
 
Join Date: Oct 2015
Posts: 1,263
Rep: 107 (Unique: 49)
Why are VRM loadline always illustrated the wrong way around? When you increase the loadline you also decrease the Vcore. So to my understanding VRM loadline should rather be illustrated like this (with the peak load on the right side and the base Vcore on the left). The higher the loadline setting the flatter the line and the lower the base Vcore, but the peak load voltage needed for stability doesn't change.
Attached Images
 
Timur Born is offline  
post #3950 of 4007 (permalink) Old 02-17-2019, 07:58 AM
New to Overclock.net
 
Falkentyne's Avatar
 
Join Date: Dec 2007
Location: Riverside
Posts: 5,416
Rep: 362 (Unique: 258)
Quote: Originally Posted by Timur Born View Post
Why are VRM loadline always illustrated the wrong way around? When you increase the loadline you also decrease the Vcore. So to my understanding VRM loadline should rather be illustrated like this (with the peak load on the right side and the base Vcore on the left). The higher the loadline setting the flatter the line and the lower the base Vcore, but the peak load voltage needed for stability doesn't change.
Because that's not how it works at all. The peak load voltage (or rather, minimum required voltage for stability) *DOES* change with maximum loadline calibration. It actually goes WAY up.

Also, the graph in bios is based on your exact voltage you are setting and the voltage used at load (although a higher loadline causes more instability at extremely heavy current due to oscillations).
You graph is talking about hypothetical situations of trying to reduce bios voltage each step to match a set target voltage. That's voodoo thinking there. Plus, reducing the voltage to the highest loadline point will cause extreme instability due to massive fluctuations causing massive dips *BELOW* the voltage being reported--you need an oscilloscope to see this.

Elmor and I have already tested this. Another recent post in the Asus forum and several in the GB section also mentioned instability trying to match on-die sense voltages, even though the temps, amps and power draw line up, the oscillations are wild. Although I wonder if they also go upwards (spikes repeatedly) also...when someone gets an oscilloscope and tests LLC8 / Ultra Extreme with FMA3/AVX small FFT Prime95, then we'll have our answers.

FYI I just finished a test for you here. This is comparing High and Turbo LLC with minimum voltage required for maximum load stability.
Prime95 29.5 build 10.
FFT's in place, 15K custom. FMA3

Minimum required vcore for stability:

LLC6 (Turbo): Bios voltage = 1.195v, MLCC (ITE 8792E) load voltage: 1.191v, VR VOUT load voltage: 1.131V. 184W CPU power (VRM), stable 2 hours. (any lower=random threads crash (usually #7 and #8 first).
LLC5 (High): Bios voltage= 1.225v, MLCC (ITE 8792E) load voltage: 1.155v, VR VOUT load voltage: 1.107V, 172W CPU power (VRM), stable 2 hours. (Temps 5C lower, so the VRM is reporting accurately).

[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

Last edited by Falkentyne; 02-17-2019 at 08:07 AM.
Falkentyne is online now  
Reply

Tags
K6 Motherboard , Msi , voltage regulator , z370 vrm , z390 , z390 vrm

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)
 


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