cancel
Showing results for 
Search instead for 
Did you mean: 

Having trouble moving from Manual Voltage to Adaptive Voltage

Zammin
Level 9
Hi everyone

First let me say that I am not an experienced overclocker by any means, I only have experience doing basic overclocking with manual voltage. Please don't crucify me if I misunderstand something or make a mistake.

With that out the way, I have a Maximus X Hero Z370 motherboard and I am currently running my 8700k at 4.8Ghz with a manual Vcore of 1.275V in BIOS (1.264V read by AIDA64, CPUID, AI suite etc) and LLC level 6. DRAM is just set to XMP 3200mhz (G.Skill Trident Z RGB 16GB). The system is stable during AIDA64 FPU stress test and can complete a Realbench stress test without error. The core voltage reading stays at 1.264V in all situations.

I have read other threads on this forum that say it's not ideal to be on LLC level 6 and full voltage all the time so I am trying to change to adaptive voltage. I have followed this guide in an attempt to do so: http://edgeup.asus.com/2017/kaby-lake-overclocking-guide/3/?_ga=2.185832041.1973363672.1518950532-88...

I understand the guide is for Kaby Lake CPUs but I am led to believe that the process remains the same for Coffee Lake and the BIOS UI is more or less the same.

The issue I am having is that after following the instructions in the guide, when i run AIDA64 FPU stress test to test stability, the core voltage continually drops and the CPU clock speed drops with it. The voltage will start at 1.264 and after starting the test it will gradually reduce all the way down to 1.00V and the clock speed to 4Ghz before I cancel the test.

I have included some screenshots of my BIOS settings that I used, please let me know if I have missed any important information and I will do my best to provide it. Any help would be appreciated. For now I have just reverted to my manual voltage settings.





5,997 Views
17 REPLIES 17

no1yak
Level 8
Try setting Multicore Enhancement to Auto. LLC to 5. LLC found in External Digital+Power Control.

Silent_Scone
Super Moderator
And AIDA isn't flagging thermal throttling whilst the test is running?
13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090

no1yak wrote:
Try setting Multicore Enhancement to Auto. LLC to 5. LLC found in External Digital+Power Control.


I could be wrong, but I remember reading that multi core enhancement has no effect when you dial in your own OC settings?

I can try LLC 5, I did try LLC 4 and the exact same thing happened though.

Silent Scone wrote:
And AIDA isn't flagging thermal throttling whilst the test is running?


Nah it's not thermal throttling, I have a custom watercooling loop with a 420mm radiatior, temps were in the 50s at the time.

Zammin wrote:
I could be wrong, but I remember reading that multi core enhancement has no effect when you dial in your own OC settings?

I can try LLC 5, I did try LLC 4 and the exact same thing happened though.



Nah it's not thermal throttling, I have a custom watercooling loop with a 420mm radiatior, temps were in the 50s at the time.



Assuming there is no throttling (use core temp to assess). Clear cmos, set the overclock again. This time, leave the multi core enhance setting at its default and don’t answer no to applying mce if prompted when enabling XMP. Doing so locks the TDP. Can’t really tell which process you went through initially, but it’s likely related to one of these. Once the TDP is locked, you have to revert to defaults to reset the condition.

Raja@ASUS wrote:
Assuming there is no throttling (use core temp to assess). Clear cmos, set the overclock again. This time, leave the multi core enhance setting at its default and don’t answer no to applying mce if prompted when enabling XMP. Doing so locks the TDP. Can’t really tell which process you went through initially, but it’s likely related to one of these. Once the TDP is locked, you have to revert to defaults to reset the condition.


Ok I'll give this a shot. Will post the results afterward. Thanks.

Zammin
Level 9
Ok so I've just carried out the above and with LLC level 4 my system can pass AIDA64 FPU stress test and Realbench Stress Test without losing clock speed. It is worth noting with LLC on auto the system crashed straight away when starting AIDA64 FPU test.

I'm seeing 1.264V idle (high performance power plan, 0.650V on balanced) however it jumps to 1.280V for a split second every so often for some reason. At full load in ADIA64 FPU test the voltage drops to 1.232V and is stable. Temps are lower than before as a result which is good:

73821

However in Realbench I see the voltage bounce around between 1.216V to 1.248V the entire time. You can see it in this video:



On a side note you can see ROG CPUID and Core temp's readings seem to flicker during the stress test, is this normal because the system is being hammered or is it a sign of instability? Realbench still passed with no errors.

73822

Forgive me for being a noob here but I'm still a little unclear as to what level of LLC is safe for my voltage setting. My understanding is that LLC reduces vdroop but can increase the voltage overshoot when coming off load. I also understand that this overshoot is so brief it's not detectable by software, which is why I'm confused about what level is safe for long term use. I got the above info from this video from Der8auer:

https://www.youtube.com/watch?v=_XGhpKHWYAg

Thanks a lot for helping me get this far, I really appreciate it.

Silent_Scone
Super Moderator
Level 4 LLC is fine. If the system is stable, no need to increase LLC further. Core temp fluctuation is perfectly normal during load transitioning whilst Realbench is being run.
13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090

Silent Scone wrote:
Level 4 LLC is fine. If the system is stable, no need to increase LLC further. Core temp fluctuation is perfectly normal during load transitioning whilst Realbench is being run.


Thanks, it's good to know LLC 4 is safe. It's very difficult for new overclockers like myself to determine since the voltage spikes are undetectable by software.

Regarding core temp and CPU-Z what I was referring to was the actual text in areas of the window flickering during Realbench. I'm not seeing any other signs of instability so I'm guessing it's just being slightly slower to update those fields while the system is being stressed to the max.

Zammin wrote:
Thanks, it's good to know LLC 4 is safe. It's very difficult for new overclockers like myself to determine since the voltage spikes are undetectable by software.

Regarding core temp and CPU-Z what I was referring to was the actual text in areas of the window flickering during Realbench. I'm not seeing any other signs of instability so I'm guessing it's just being slightly slower to update those fields while the system is being stressed to the max.


That's just where the interface is refreshing. If it didn't, you wouldn't be able to tell very much from it 😉
13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090