cancel
Showing results for 
Search instead for 
Did you mean: 

R5E VCCIN LLC 7 & 8 "measured".

Raja
Level 13
The following table shows VCCIN level 7&8 LLC idle and load voltages measured with an oscilloscope. Transient effects are also shown. Level 7 or 8 is recommended for 24/7 use.

47965



All other DIGI+ settings were at default. Load used to make these measurements was 260~300W.
45,505 Views
16 REPLIES 16

GoNz0-
Level 10
pros and cons to level 9 as my system set itself to that via auto?

Raja
Level 13
With level 9 the voltage will be higher than what is set in UEFI when the system has settled under load. The elevation is set so that the peak voltage sag does not appear to be far away from the applied VID (for people that prefer seeing that type of stuff in software and on their DMMs).

My LLC is set to 5 by default.
Should I set it to 7-8?
NOTE: 5820k OCed to 4GHz @ 1.25v / 3GHz CPU Cache (stock cache voltage)

Raja
Level 13
You can use any level you wish.

Raja
Level 13
Some captures below. Probe tips placed directly across MLC capacitor at back of socket, with short ground lead to limit noise pickup, and bandwidth limiting enabled on scope. Default VRM switching frequency. Load is around 300W generated via software:

48770
LLC 7 Idle to load transition.


48771
LLC7 load to idle transition.




LLC 9 is more strenuous on the VRM as the voltage is increased under load. The VRM handles this fairly well however and keeps overshoot fairly tight (50mv over load voltage level) - it's set this way to appease people that think voltage should not move when faced with a load. These images serve to show that it does move - the laws of physics rule. If running LLC 9, one just needs to bear in mind that the voltage will be higher than what is set under full load and will peak around 0.1V over what is set in UEFI briefly when transitioning from load to idle. The overshoot period is very brief - says a lot about how good the VRM solutions on these boards are. 🙂



48772
LLC9 idle to load transition.


48773
LLC9 load to idle transition.



I've posted these images to save people from getting into arguments over what to run. Pick what suits your mindset. If you care about the max voltage your CPU sees, just offset the max you're willing to feed the CPU by the peak level. So if your max is 1.95V for VCCIN, you would use 1.85V in UEFI with LLC 9 as an example.

Do not ask me for more measurements as I have taken the setup apart and am loathe to go through with it again! I took the measurements above six weeks ago, not today. Might take a look at other things in the future on other products, but won't be revisiting this.

Enjoy!

Raja@ASUS wrote:
Some captures below. Probe tips placed directly across MLC capacitor at back of socket with short ground lead to limit noise pickup and bandwidth limiting enabled on scope. Default VRM switching frequency. Load is around 300W generated via software:

48770
LLC 7 Idle to load transition.


48771
LLC7 load to idle transition.




LLC 9 is more strenuous on the VRM as the voltage is increased under load. The VRM handles this fairly well however and keeps overshoot fairly tight (50mv over load voltage level) - it's set this way to appease people that think voltage should not move when faced with a load. These images serve to show that it does move - the laws of physics rule. If running LLC 9, one just needs to bear in mind that the voltage will be higher than what is set under full load and will peak around 0.1V over what is set in UEFI briefly when transitioning from load to idle. The overshoot period is very brief - says a lot about how good the VRM solutions on these boards are. 🙂



48772
LLC9 idle to load transition.


48773
LLC9 load to idle transition.



I've posted these images to save people from getting into arguments over what to run. Pick what suits your mindset. If you care about the max voltage your CPU sees, Jjust offset the max you're willing to feed the CPU by the peak level. So if your max is 1.95V for VCCIN, you would use 1.85V in UEFI with LLC 9 as an example.

Do not ask me for more measurements as I have taken the setup apart and am loathe to go through with it again! I took the measurements above six weeks ago, not today. Might take a look at other things in the future on other products, but won't be revisiting this.

Enjoy!


Great stuff Raja! The VRM used on the RVE are very tight at a 50mV transient overshoot! Very helpful.
Thanks! 🙂

Raja
Level 13
Yes its tight - under 5us settling time even when releasing a load well outside stock spec.

I have 2 questions, one for RAJA and the other one for anyone here that wants to give an advice on the picture that I'm going to ask if I'm allow to post on this sticky. At RAJA can I post here 4 pictures of my AIDA bench of two different LLC.
"I'm not an OCer just looking a good MENTOR". UyqtCojoXecQlo.". 😄