[SOLVED] Ryzen 5 3600 voltage

redbioniclem

Reputable
Jul 3, 2019
51
7
4,535
My build:

https://pcpartpicker.com/user/redbioniclem/saved/#view=8bvdmG

Hi, I am using the Ryzen 5 3600 but I am not planning on overclocking it and it has the stock cooler with the Arctic mx4(2019) thermal paste.

The temperature at 1.4 voltage (stock) is 50 C- 60 C in an idle state.

That's why I want to know if I should undervolt the CPU and what voltage I should use? I tried 1.2125 and the temperature is around 45 C - 55 C. And how do I make that voltage to be applied when I boot the PC and not from the Ryzen master app.

Also, should I leave the fan at 100% all the time?
I don't care about the sound but, is it OK to do that?

Please help!
 
Last edited by a moderator:
Solution
If you are okay with sacrificing some performance then its fine. I would probably go into AMD Overclocking> Precision Boost Overdrive>Advanced> and set a 75c temp limit rather than undervolt.

redbioniclem

Reputable
Jul 3, 2019
51
7
4,535
I think around 1.2-1.25V should be fine for stock operations.
Early adopter problems, unfortunately - where early release BIOS' revisions do have some kinks to be worked out.

Might be worth checking out this video, as there's more than just the basic voltage at play:
View: https://www.youtube.com/watch?v=Ssuqhyqah2k
Would you know how to set the voltage so the pc boots with that voltage?
Also, I updated the post description and included my build.
 

daysyang

Distinguished
Mar 14, 2012
122
1
18,695
Have to set it in your bios, press delete while computer is turning on. Go into the advance mode of the bios, and go to overclock settings. This may look different depending on mobo manufacture, asus: ai tweeker, giga: m.i.t, msi: oc from there just go to cpu core voltage and set your desire voltage. Default is auto, you can try setting it manual or preferably adaptive, save and exit.
 
  • Like
Reactions: DMAN999

rigg42

Respectable
Oct 17, 2018
639
233
2,390
Don't under-volt as it will kill your performance. I had my 3700x stable with a 200mv under-volt. Clocks, thermals, and power usage looked amazing in stress tests. Then I ran cinebench and realized my performance was completely neutered. Same story on my 3600 and 3600x.
 

redbioniclem

Reputable
Jul 3, 2019
51
7
4,535
Don't under-volt as it will kill your performance. I had my 3700x stable with a 200mv under-volt and clocks, thermals, and power usage looked amazing. Then I ran cinebench and realized my performance was completely neutered. Same story on my 3600 and 3600x.

I don't have much experience since this is my first pc build that I did just adding parts to an older built pc that I bought. how can I test and understand if my performance is not great, I am thinking about undevolting because since I have the CPU stock cooler I was worried about the temperatures.
 

rigg42

Respectable
Oct 17, 2018
639
233
2,390
I don't have much experience since this is my first pc build that I did just adding parts to an older built pc that I bought. how can I test and understand if my performance is not great, I am thinking about undevolting because since I have the CPU stock cooler I was worried about the temperatures.
Replace the cooler. I've spent the last 5 days playing with this stuff on 3 differnt 3k CPUs. You can run your games and benchmarks to compare or you can take my word that undervolting is a bad idea.
 
  • Like
Reactions: DMAN999

daysyang

Distinguished
Mar 14, 2012
122
1
18,695
Before changing voltage try testing it with cinabench and just playing games for 30min-1hr. You can compare cinabench score with others online. If your games have frame rate turn it on. Check your voltage and temps, make notes, then you can go back and make changes if necessary. Idle at 50-60c, try removing/reapplying paste and remount cpu.
 

redbioniclem

Reputable
Jul 3, 2019
51
7
4,535
Replace the cooler. I've spent the last 5 days playing with this stuff on 3 differnt 3k CPUs. You can run your games and benchmarks to compare or you can take my word that undervolting is a bad idea.

I cannot buy a new cooler right now, so for now should I just leave the voltage as it is ? even with higher temperatures?
 

rigg42

Respectable
Oct 17, 2018
639
233
2,390
If you are okay with sacrificing some performance then its fine. I would probably go into AMD Overclocking> Precision Boost Overdrive>Advanced> and set a 75c temp limit rather than undervolt.
 
Solution

Barty1884

Retired Moderator
If you are okay with sacrificing some performance then its fine. I would probably go into AMD Overclocking> Precision Boost Overdrive>Advanced> and set a 75c temp limit rather than undervolt.

You'd rather leave it at 1.4V (+/-) and throttle it &/or have the fan get fairly loud, vs a simply voltage tweak?

It's worth checking though OP, run Cinebench, log your results. Then manually set the voltage to 1.3V and re-run Cinebench. You won't damage anything, and can easily change it back.
That way you can verify @rigg42 's anecdotal evidence of 1, that undervolting the CPU (yet still providing plenty voltage) will hinder the CPU. It might, but I'd be surprised.
 

rigg42

Respectable
Oct 17, 2018
639
233
2,390
You'd rather leave it at 1.4V (+/-) and throttle it &/or have the fan get fairly loud, vs a simply voltage tweak?

It's worth checking though OP, run Cinebench, log your results. Then manually set the voltage to 1.3V and re-run Cinebench. You won't damage anything, and can easily change it back.
That way you can verify @rigg42 's anecdotal evidence of 1, that undervolting the CPU (yet still providing plenty voltage) will hinder the CPU. It might, but I'd be surprised.
It only runs at that voltage at idle and low load. There is hardly any current going through the cpu in that state and its completely safe. The voltage shifts dynamically based on load. Fully loaded it will run in the low 1.3's. Setting the temp limit will lower the voltage and temp limits (obviously) when heavily loaded but won't kill single threaded performance like undervolting will.
 

rigg42

Respectable
Oct 17, 2018
639
233
2,390
You'd rather leave it at 1.4V (+/-) and throttle it &/or have the fan get fairly loud, vs a simply voltage tweak?

It's worth checking though OP, run Cinebench, log your results. Then manually set the voltage to 1.3V and re-run Cinebench. You won't damage anything, and can easily change it back.
That way you can verify @rigg42 's anecdotal evidence of 1, that undervolting the CPU (yet still providing plenty voltage) will hinder the CPU. It might, but I'd be surprised.
I have 3 of these Ryzen 3k CPU's.
 

Barty1884

Retired Moderator
It only runs at that voltage at idle and low load. There is hardly any current going through the cpu in that state and its completely safe. The voltage shifts dynamically based on load. Fully loaded it will run in the low 1.3's. Setting the temp limit will lower the voltage and temp limits (obviously) when heavily loaded but won't kill single threaded performance like undervolting will.

Except there's absolutely no need for ~1.4V full stop, idle/low or otherwise.

Without seeing proof of it running in the ~1.3V range under full load, forgive me, but I highly doubt that.

I have 3 of these Ryzen 3k CPU's.

Sorry, anecdotal evidence of 3.
 

Barty1884

Retired Moderator
Your results are showing 31-44 seconds? I'd be inclined to suspect you have an issue that would appear in Prime eventually.

As for the Cinebench score..... what kind of voltage went along with that? You undervolted, sure.... but I don't see anything to verify what's running.
 

rigg42

Respectable
Oct 17, 2018
639
233
2,390
Your results are showing 31-44 seconds? I'd be inclined to suspect you have an issue that would appear in Prime eventually.

As for the Cinebench score..... what kind of voltage went along with that? You undervolted, sure.... but I don't see anything to verify what's running.
I'm confused. You asked me to show the difference between idle and full voltage behavior. I don't understand why that takes more than 30 seconds to show you. You can see the VID went from 1.5 at idle to 1.313 at full load. I'm sure the idle voltage varies from Mobo to mobo and CPU SKUs but the cpu will regulate its voltage based on load. I think I've demonstrated this clearly. That cinebench run was at 1.3 volts. I don't know how you expect me to prove that to you or why you are taking such a combative and condescending tone toward me. What incentive is there for me lie about the Cinebench score and voltage?
 

Barty1884

Retired Moderator
I apologize if you feel I'm coming across as combative or condescending, that's not my intent.

While it's branching a little off topic for this specific thread, my curiosity is peaked - hence the question(s).
I do see that the VID went down, yes. All I was trying to convey is that while yes, it would suggest that once the CPU hits full load, the VID dropped to 1.313, ~30 seconds doesn't demonstrate that to be true (nor prove it to be false, of course).
Equally, if there is an underlying instability in your system(s), 30 seconds may not be sufficient to find same. It doesn't make a whole lot of sense to me that a CPU can auto regulate it's voltage to 1.313V and yet manually setting to the same would cause a problem.
To be clear, I'm not saying that it can't or isn't causing a problem, just that it doesn't appear to make any sense on the face of it.

I'm not implying there's incentive for you to lie about the Cinebench result, it's clearly a low score. What I am curious about though, is whether that's truly due to the undervolting or whether there's an underlying problem - potentially the underlying problem that Prime was yet to find after 30 seconds.
 

rigg42

Respectable
Oct 17, 2018
639
233
2,390
I apologize if you feel I'm coming across as combative or condescending, that's not my intent.

While it's branching a little off topic for this specific thread, my curiosity is peaked - hence the question(s).
I do see that the VID went down, yes. All I was trying to convey is that while yes, it would suggest that once the CPU hits full load, the VID dropped to 1.313, ~30 seconds doesn't demonstrate that to be true (nor prove it to be false, of course).
Equally, if there is an underlying instability in your system(s), 30 seconds may not be sufficient to find same. It doesn't make a whole lot of sense to me that a CPU can auto regulate it's voltage to 1.313V and yet manually setting to the same would cause a problem.
To be clear, I'm not saying that it can't or isn't causing a problem, just that it doesn't appear to make any sense on the face of it.

I'm not implying there's incentive for you to lie about the Cinebench result, it's clearly a low score. What I am curious about though, is whether that's truly due to the undervolting or whether there's an underlying problem - potentially the underlying problem that Prime was yet to find after 30 seconds.
These are the stock settings. I've fully stress tested this system. An hour off small FFT, 10 consecutive IBT runs, 8 hours of realbench, multiple superposition runs, and a stock memtest run. What happens when you lock a fixed voltage is it prevents XFR and PB from fully boosting the cores. It won't affect the multi core very much at fixed voltage unless you lower it a lot. even locking at 1.35 did lower my cinebench multicore about 100 pts.
 

Barty1884

Retired Moderator
These are the stock settings. I've fully stress tested this system. An hour off small FFT, 10 consecutive IBT runs, 8 hours of realbench, multiple superposition runs, and a stock memtest run. What happens when you lock a fixed voltage is it prevents XFR and PB from fully boosting the cores. It won't affect the multi core very much at fixed voltage unless you lower it a lot. even locking at 1.35 did lower my cinebench multicore about 100 pts.

I appreciate the theory behind it, but with that rationale, a single core stress test needs the upper end of that "stock" voltage spectrum?
So 4.5GHz (for example) with XFR/PB on a single core needs 1.5V, but 4.0-4.1GHz across all cores (also boosted by XFR/PB) needs ~1.3V?
Again, not calling you out on anything, or trying to be codescending or antagonistic, I just can't wrap my head around it.

This is not unlike the early 300 series BIOS' though, where unnecessary voltage was pushed through the CPU. It did get better over time, with performance improving IIRC, so perhaps a BIOS update will fix the issue in time.

Anyway, I've derailed this thread enough.
 

rigg42

Respectable
Oct 17, 2018
639
233
2,390
If you read the AMD post i linked above you'll see that that sky high voltages at idle are because the CPU isn't really idle. The monitoring software is pinging it with a tiny load and causing the boost algorithm to fire to full boost (and the voltage it needs to achieve it). As i mentioned earlier these voltages are fine in bursts at light load because there isn't enough current to damage the CPU. The CPU is doing exactly what its supposed to do. The OP's temperature problem is because of the crappy stealth cooler.
 

rigg42

Respectable
Oct 17, 2018
639
233
2,390
I appreciate the theory behind it, but with that rationale, a single core stress test needs the upper end of that "stock" voltage spectrum?
So 4.5GHz (for example) with XFR/PB on a single core needs 1.5V, but 4.0-4.1GHz across all cores (also boosted by XFR/PB) needs ~1.3V?
Again, not calling you out on anything, or trying to be codescending or antagonistic, I just can't wrap my head around it.

This is not unlike the early 300 series BIOS' though, where unnecessary voltage was pushed through the CPU. It did get better over time, with performance improving IIRC, so perhaps a BIOS update will fix the issue in time.

Anyway, I've derailed this thread enough.
Yes that's precisely what I'm getting at. The Bios is fine. Its doing what its supposed to do. The problem is HWI64 and how it interacts with the CPU on the ryzen balanced power plan. The cores aren't idle when they're at 4.4ghz and 1.5 v. I just downloaded CPUz and my voltage was hovering around .9 with nothing else running. HWI64 shows the same if I turn on the windows balanced power plan.

On a side note the SOC voltage stuff that J2C was talking about in that video is a legit issue with the Asus bios and fast memory kits. Even on Intel. My Max XI hero did the same thing to VCCIO and VCCSA when i threw a 3600 cl16 kit in it. They really need to fix that. That's a recipe for burned up memory controllers.
 
Last edited: