ryzen 3900x voltage going up to 4.6~4.7 after hours of activity

[1984]

Checked it using Ryzen master. All settings are on auto in BIOS. Mobo is asus crosshair hero 8. Also to note, on Auto precision boost overdrive is off and ryzen master just shows default OC instead,

Normally on default OC when you launch pc it stays at about 1.1 and even if u power up all cores for program it doesnt go above 1.3

PBO on the other hand does go to 1.46 range but it does so almost immediately.

Anyway, just wondering what's going on

[Snow]

When you manually OC you adjust general voltage and general clocks. PBO, on the other hand, can take one-two cores your PC currently uses (it's never truly idling, there always is some stuff going on) and applies extra frequency/voltage to those cores specifically, living all other cores out of this. From how I see it - it's doing exactly what it should, giving more juice to cores that need that in exact moment of time.

[chiefputsilao✖️ping]

are your sure the VOLTAGE is 4.6 - 4.7 ?
i think it is the boost FREQUENCY.

[r.linder]

It's normal for voltages to reach 1.4~1.5. There's nothing unsafe about it because it caps off at 1.5.

[UserNotFound]

One way to get around it is by OC'ing it and adjusting CPU Voltage yourself to lowest stable at your OC'ed speed. I have mine at 4.475Ghz at 1.35V, stable enough to 3DMark and the games I play, Metro Exodus, Gears 5, UT3.

[SeriousCCIE]

Is it true that when things are set to auto as described, there is a normal default OC voltage?

I am greatly inclined to agree with Snow and Escorve because those have also been my experiences with CPU voltages in general.

Based on past experiences.. . I would expect that cpu voltage to be variable depending on the speed, load, and physical core utilization (not to mention occupation of the L2 and L3 with held data and access to and from it, etc).

That all means that I think the viewed voltatges would be normal, but I have to state I don't have direct experience with the 3900x and can't say that it is or isn't--just that it'd be odd that the values are absolute and capped like that. I'd think such capping is possible, but not desired in a high performance mode...

[r.linder]

引用自 SeriousCCIE：

Is it true that when things are set to auto as described, there is a normal default OC voltage?

I am greatly inclined to agree with Snow and Escorve because those have also been my experiences with CPU voltages in general.

Based on past experiences.. . I would expect that cpu voltage to be variable depending on the speed, load, and physical core utilization (not to mention occupation of the L2 and L3 with held data and access to and from it, etc).

That all means that I think the viewed voltatges would be normal, but I have to state I don't have direct experience with the 3900x and can't say that it is or isn't--just that it'd be odd that the values are absolute and capped like that. I'd think such capping is possible, but not desired in a high performance mode...

Voltages are also set per core, not across all cores. (1 core can be 1.41~1.5 while another can be 1.35~1.4, with different clocks per core) Idle voltages are still a bit high, but it's not really the case for load... we're talking about a 12-core 24-thread beast of a processor that's essentially a Threadripper CPU squeezed into an AM4 package.

AMD has auto/PBO set to cap off at 1.5v via a software limitation set in the BIOS, as far as I can understand. It makes sense because letting it go as high as it thinks it can handle can kill the chip.

Board manufacturers are in the process of releasing AGESA 1.0.0.3 ABBA which is supposed to fix most of the issues with idle voltage. X570 boards mostly have it already, but I'm still waiting for my X470 AORUS GAMING 7 WIFI to get it. Gigabyte needs to show their previously top end board users some love.

[SeriousCCIE]

Thanks for the detailed reply; I knew AMD recently resolved the idle voltages and turbo boost frequency issues about being unable to stay, or even reach, at the peak.

And that also not a lot of OEMs have packaged the update for mainstream use quite yet. I imagine the microcode is probably accessible otherwise but then... there are not mature tools for that quite yet in the '3rd party' market quite yet. (in time, maybe...)

Which leads me to a question that might veer off topic, but I think it is relevant.

I'd expect that the motherboard firmware (bios/cmos, etc) would have a thermal regulation check--like a bypass or something to turn off for those users that are going to kill their CPU if using conventional cooling, but they aren't using conventional cooling.

You mentioned LN2 earlier--no normal person uses that for day to day usage, but water cooling certainly is. Does AMD allow the 1.5v voltage limitation to be exceeded--whether or no it can determine what the cooling method is via sensors?

Just trying to get an idea as to sustained performance options as opposed to peak values that can't be reached without regular shipments of hazardous materials (which LN2 is classified as in my neck of the woods...LN2 is too expensive to use regularly even if it had a subscription with free shipping!)

I understand that AMD has taken the approach of making things more efficient and easier to reach, but I'd still expect (or hope maybe) that there are some silicon lottery winners that could do more with creative cooling.

[r.linder]

引用自 SeriousCCIE：

Thanks for the detailed reply; I knew AMD recently resolved the idle voltages and turbo boost frequency issues about being unable to stay, or even reach, at the peak.

And that also not a lot of OEMs have packaged the update for mainstream use quite yet. I imagine the microcode is probably accessible otherwise but then... there are not mature tools for that quite yet in the '3rd party' market quite yet. (in time, maybe...)

Which leads me to a question that might veer off topic, but I think it is relevant.

I'd expect that the motherboard firmware (bios/cmos, etc) would have a thermal regulation check--like a bypass or something to turn off for those users that are going to kill their CPU if using conventional cooling, but they aren't using conventional cooling.

You mentioned LN2 earlier--no normal person uses that for day to day usage, but water cooling certainly is. Does AMD allow the 1.5v voltage limitation to be exceeded--whether or no it can determine what the cooling method is via sensors?

Just trying to get an idea as to sustained performance options as opposed to peak values that can't be reached without regular shipments of hazardous materials (which LN2 is classified as in my neck of the woods...LN2 is too expensive to use regularly even if it had a subscription with free shipping!)

I understand that AMD has taken the approach of making things more efficient and easier to reach, but I'd still expect (or hope maybe) that there are some silicon lottery winners that could do more with creative cooling.

Pretty sure you can go well beyond motherboard set limits, but it's still not recommended outside of LN2.