cancel
Showing results for 
Search instead for 
Did you mean: 

Maximus Formula XI Z390 - VDroop

Wih__Glah
Level 7
Hi!

I am seeing a significant VDroop on the VCore of my 9900k.

I set the Vcore manually to 1.22 in the BIOS. During normal activities it drops to 1.18v and during AVX it drops to 1.14v

All cores at 5GHz, AVX offset 1. LLC set to 6,7 or 8 makes no difference.

Power limits have been maxed as per Debaur's guide on 8700K and Intel ETU states I am not power throttling.

Any ideas?
14,242 Views
29 REPLIES 29

PerpetualCycle
Level 13
yes

ROG Dark Hero Z790 | 13900KS @5.7 GHz | g.skill 2x48GB 6800 MT/s | ROG Strix 4070 Ti | EK Nucleus 360 Dark | 6TB SSD/nvme, 16TB external HDD | 2x 1440p | B&W 606 S3 speakers, Martin Sub, Audiolab 7000A amp| Fractal North XL case

Yes, the other brands do not report as accurately compared to the new XI series ROG boards. So what we are seeing now is closer to the real voltage and vdroop.

It makes sense as well. When I compare the lowest stable reported voltage on the Hero XI for a given CPU to the lowest stable reported voltage on "another brand" it appears that the Hero XI required stable voltage is much lower. But that is not what's happening. The other board is simply inferring a less accurate higher voltage. In order to get the other brand to report the same loaded voltage as the Hero XI you need to drop voltage down about another 100mv at LLC 6. However this results in unstable voltage levels for the chip.

Bottom line: the cpu is actually seeing the same voltage levels on both boards at set voltage levels at LLC 6, but the Hero XI reports a lower more accurate voltage. Now, this may encourage some users to crank up the voltage more, or push the LLC to 7 or 8, but I would caution against that. Some vdroop is good. Instead just be aware that your chip will appear to run stable on lower voltages with more vdroop than you might have been accustomed to seeing.

mdzcpa wrote:
Yes, the other brands do not report as accurately compared to the new XI series ROG boards. So what we are seeing now is closer to the real voltage and vdroop.

It makes sense as well. When I compare the lowest stable reported voltage on the Hero XI for a given CPU to the lowest stable reported voltage on "another brand" it appears that the Hero XI required stable voltage is much lower. But that is not what's happening. The other board is simply inferring a less accurate higher voltage. In order to get the other brand to report the same loaded voltage as the Hero XI you need to drop voltage down about another 100mv at LLC 6. However this results in unstable voltage levels for the chip.

Bottom line: the cpu is actually seeing the same voltage levels on both boards at set voltage levels at LLC 6, but the Hero XI reports a lower more accurate voltage. Now, this may encourage some users to crank up the voltage more, or push the LLC to 7 or 8, but I would caution against that. Some vdroop is good. Instead just be aware that your chip will appear to run stable on lower voltages with more vdroop than you might have been accustomed to seeing.


Is there anything wrong with running something like this?

1.42v bios, 1.39v actual voltage in windows idle, 1.33v load LLC6

Jaz11 wrote:
Is there anything wrong with running something like this?

1.42v bios, 1.39v actual voltage in windows idle, 1.33v load LLC6


If you are running stable with good temperatures than you are fine. However, once you have determined those values are stable, you can always experiment with lower voltages and test for stability. The lowest required voltage is always best.

New bios just uploaded today. I will give it a go after work. Patch notes look promising but i doubt it will drastically change anything

"Version 0602
2018/10/318.52 MBytes
ROG MAXIMUS XI FORMULA BIOS 0602
- Improved system stability and performance.
- Updated Microcode for 8-core CPU.
- Improved DRAM stability.
- Fixed USB device issue."

Wih__Glah
Level 7
So I have managed to get my Vcore stable under normal loads with a -ve offset of 0.06v with Auto settings
However under AVX loads it adds another 100mV and temperatures shoot up ridiculously, so I am back to manual setting

Currently :

VCORE BIOS set = 1.225V
VCORE HW Monitor Idle = 1.208V
VCORE HW Monitor Load = 1.137V
VCORE HW Monitor AVX load = 1.121V
LLC 6

That's 71mV of vdroop. 87mV under AVX.

Is it safe to go to LLC 7 permanently?


Latest BIOS BTW

Wih\'Glah wrote:
So I have managed to get my Vcore stable under normal loads with a -ve offset of 0.06v with Auto settings
However under AVX loads it adds another 100mV and temperatures shoot up ridiculously, so I am back to manual setting

Currently :

VCORE BIOS set = 1.225V
VCORE HW Monitor Idle = 1.208V
VCORE HW Monitor Load = 1.137V
VCORE HW Monitor AVX load = 1.121V
LLC 6

That's 71mV of vdroop. 87mV under AVX.

Is it safe to go to LLC 7 permanently?


Latest BIOS BTW


I agree with Tostito, I wouldn't recommend higher than LLC 7. Some vdroop is good to protect against overshoot. In my case, I still run LLC 6 and deal with the 100mv droop. (I set 1.270v in the BIOS and measure 1.190v with HWM or CPUZ). Chip seems very stable to I'm happy.

tostitobandito
Level 7
LLC7 is fine. You only really need to be cautious about LLC8, and not at your voltages regardless. Even if you overshoot for a split second it won't be enough voltage to hurt anything. If your vcore is 1.40v or something it's another story and you want to be more careful with aggressive LLC.

DarksideEE7
Level 7
I experience a fairly significant (0.098V) Vdroop on the Maximus XI Formula using LLC Auto (ends up being 6) with 1.36V adaptive voltage in UEFI @5.1 GHz. I would love to hook up an oscilloscope to this board, however it doesn't have the measurement pads like other boards in the past. Does anyone know where to measure Vcore with a DMM or oscilloscope by any chance?
77422
I'm stable, so I'm not that concerned about Vdroop, though I would like to minimize it. If Elmor's post on OCN is accurate (I'm sure it is), then the SIO is close enough to the CPU-die sense for my comfort. 12mV is not bad at all. I'm going to try setting LLC to 6 manually, however it would be nice to hook up a scope an see what kind of transient overshoot we get at each LLC level.

I also wonder about the CPU VRM Switching Frequency setting. I have it set to auto. According to the description, increasing it should allow us better transient response during transition to load. It says the max value is 500 kHz. Anyone tried playing with this? The ability to hook up an oscilloscope and get the Vcore waveform during loading would definitely help when playing with this setting...

Also, does anyone know if the Maximus XI Formula has the same VRMs as the XI Hero?

EDIT:

After testing, I see that I get the same Vdroop with LLC6 set manually instead of set to auto (which ends up being LLC 6).

I don't think i have this vdroop problem. After multiple crashes and BSODs, i have finally settled on the following stable settings:

Asus Code XI - 9700K - 5.0 GHz - AVX offset 0 - Adaptive - Additional turbo voltage 1.29 - under load 1.288 - LLC5 - BIOS ver. 1401 - Temps max 74 under load on hottest core (Core #2) - air cooled NH-D15