Question Undervolt for my CPU some how increases voltage

Sep 29, 2021
30
3
35
0
I have a AMD 3600(no overclock) (artic 34 duo esports cooler)
Basically, a few days a go i made a post because i was concerned about my CPU voltage. It was genreally 1.39-1.4 when playing games. Although temp was a problem maximum being 60c, averaging 55c. One of the suggestions was to undervolt my CPU, so i looked on youtube followed a tutorial and did so.
These are the steps I did.
Download cinebench and do the multicore and single core tests and get results. I did that and got 8868 points in multi core and 1190 in single core. 70c in multi core and I really remember but around 60c in single core. I didnt look at voltage cus in the video it didn't but this was before the undervolt anyways so it didnt matter to me.
Then I did the undervolt by going to BIOS, OC Settings. CPU Core voltage set to offset mode, press '-' the offset is taking away , and then I inputted 0.0750v to take away.
Now go to the tests again, I got 9170 in the multi core and I couldnt be asked to do the single core. Multicore temp was 65c. Now I did this to lower voltage not nessacrly temps since my temps are already good so this time i looked at voltage using Ryzen master and it was 1.41 sometimes geting into 1.42. I did further testings played a game called HoldFast in which my CPU is usually 1.39v it was not 1.41 - 1.42v. Although the temps were lowered from 55c normally to now 49c.
I wanted to lower voltage so my CPU lasts longer and it hasn't done so infact it has done the opposite. Any help or suggestions?
 

tennis2

Judicious
1.4V isn't harmful. Neither is 70C.

Here's what a voltage offset does: The CPU operates along a pre-defined frequency/voltage curve. At any given frequency, there is a corresponding voltage. By applying a voltage offset it just tells the mobo to apply that much more/less voltage for any given frequency point on the curve. An offset voltage does not CAP the voltage that will be used. That limit is handled by the CPU boost algorithms (PBO, etc).

I did that and got 8868 points in multi core 70c in multi core
Now go to the tests again, I got 9170 in the multi core. Multicore temp was 65c.
Although the temps were lowered from 55c normally to now 49c.
So your undervolt allowed the CPU to boost frequency and improve your scores, while also decreasing temp by 5C. Sounds like a win to me.

Now I did this to lower voltage not nessacrly temps since my temps are already good so this time i looked at voltage using Ryzen master and it was 1.41 sometimes geting into 1.42. I did further testings played a game called HoldFast in which my CPU is usually 1.39v it was not 1.41 - 1.42v.
What exactly are you using to monitor voltage? When are you monitoring it? For example, you're likely to see higher voltage with 1-2 cores loaded compared to all cores loaded since the 1&2 core frequency is higher than all-core.
For example, my 5600G will do 4GHz all-core @ 1.2V, but 4.6GHz single-core @ 1.44V.
 
Last edited:
Sep 29, 2021
30
3
35
0
1.4V isn't harmful. Neither is 70C.

Here's what a voltage offset does: The CPU operates along a pre-defined frequency/voltage curve. At any given frequency, there is a corresponding voltage. By applying a voltage offset it just tells the mobo to apply that much more/less voltage for any given frequency point on the curve. An offset voltage does not CAP the voltage that will be used. That limit is handled by the CPU boost algorithms (PBO, etc).


So your undervolt allowed the CPU to boost frequency and improve your scores, while also decreasing temp by 5C. Sounds like a win to me.


What exactly are you using to monitor voltage? When are you monitoring it? For example, you're likely to see higher voltage with 1-2 cores loaded compared to all cores loaded since the 1&2 core frequency is higher than all-core.
For example, my 5600G will do 4GHz all-core @ 1.2V, but 4.6GHz single-core @ 1.44V.
I was using Ryzen master to monitor volts, im now using HWINFO but idk which setting to look at.
 

tennis2

Judicious
I was using Ryzen master to monitor volts, im now using HWINFO but idk which setting to look at.
Ultimately, if you're just scared to see 1.4V (justified or not) it probably doesn't matter. But like I said, the 1.4V is probably only happening every now and then when you get a frequency spike to the single-core boost frequency. I prefer using Prime95 to test since it just loads the set # of cores 100% indefinitely so it makes monitoring frequency vs voltage easier.

In HWinfo for undervolting, you want to monitor "effective" frequency on each core. That's going to show whether you're experiencing clock stretching (effective frequency is lower than the "general"(?) frequency) as you undervolt further and further. Clock stretching will decrease the performance.
 
Last edited:

ASK THE COMMUNITY

TRENDING THREADS