Question Ryzen 3000 Voltages when dealing with manual overclocking

[Redneck5439]

I think by this time we all know that XFR and PBO auto use voltage very efficiently with Ryzen 3000. We have all seen where voltage spikes up to 1.5V are perfectly fine when the processor is boosting as it is a very quick spike and its simply normal operation. The question I want to pose to the experts is what is the safe "static" voltage for Ryzen 3000 when manually overclocking?

I have seen 1.325V thrown around, but with no technical data behind it, and AMD has not released any official information for overclocking. I have overclocked several Ryzen systems now but will not exceed 1.36V, and on my own personal rig won't exceed 1.375V.

At 1.375V I have managed a 4.4Ghz all core overclock which I further overclocked in Ryzen Master so my fastest CCX is running at 4.425Ghz and the other is running at the bios set 4.4Ghz. I know if I up the voltage to 1.39V I am stable at 4.45Ghz all core in bios (and may be able to get a little more with CCX overcloking), but I am uncomfortable running at that without knowing for sure what the max Vcore for Ryzen 3000 truly is.

Currently I use a voltage offset so my processor doesn't have the full 1.375V load all the time. My voltage will scale from 0.5 to 1.375V depending on the load and my cores are configured so they enter "sleep" when not under load.

What do the experts out there have to say about 7nm Ryzen 3000 for sustained Vcore under load?

 

[InvalidError]

AMD has not released any official information for overclocking.

Of course it isn't releasing any information on overclocking. While AMD isn't going out of its way to prevent it, it isn't officially supporting it either. You do it at your own risk with no guarantees expressed or implied from AMD.

The reason why Vcore on modern CPUs is so much higher than chips from only a few years ago despite process tech going smaller which used to bring lower Vcore is because modern chips use internal LDO (low-drop-out) linear regulators to feed cleaner voltage to logic (noisy power makes logic take longer to settle, increases power draw from spending more time in in-between states and reduces attainable clock frequencies) and those need 100-300mV of overhead to do their job, so a bunch of voltages are now nominally ~200mV higher than they used to be to compensate and also as a buffer for the VRM's transient response.

1.5V is a common idle/light-load voltage because it gives the LDOs the most headroom for peak momentary boost.

 

[coolcole01]

I think by this time we all know that XFR and PBO auto use voltage very efficiently with Ryzen 3000. We have all seen where voltage spikes up to 1.5V are perfectly fine when the processor is boosting as it is a very quick spike and its simply normal operation. The question I want to pose to the experts is what is the safe "static" voltage for Ryzen 3000 when manually overclocking?

I have seen 1.325V thrown around, but with no technical data behind it, and AMD has not released any official information for overclocking. I have overclocked several Ryzen systems now but will not exceed 1.36V, and on my own personal rig won't exceed 1.375V.

At 1.375V I have managed a 4.4Ghz all core overclock which I further overclocked in Ryzen Master so my fastest CCX is running at 4.425Ghz and the other is running at the bios set 4.4Ghz. I know if I up the voltage to 1.39V I am stable at 4.45Ghz all core in bios (and may be able to get a little more with CCX overcloking), but I am uncomfortable running at that without knowing for sure what the max Vcore for Ryzen 3000 truly is.

Currently I use a voltage offset so my processor doesn't have the full 1.375V load all the time. My voltage will scale from 0.5 to 1.375V depending on the load and my cores are configured so they enter "sleep" when not under load.

What do the experts out there have to say about 7nm Ryzen 3000 for sustained Vcore under load?

That's awesome on the 4.4 all core overclock, I am barley able to hit over 4.3. I have been running it pretty sustained at 1.387 and have yet to see anything negative voltage related.

 

[Redneck5439]

Of course it isn't releasing any information on overclocking. While AMD isn't going out of its way to prevent it, it isn't officially supporting it either. You do it at your own risk with no guarantees expressed or implied from AMD.

The reason why Vcore on modern CPUs is so much higher than chips from only a few years ago despite process tech going smaller which used to bring lower Vcore is because modern chips use internal LDO (low-drop-out) linear regulators to feed cleaner voltage to logic (noisy power makes logic take longer to settle, increases power draw from spending more time in in-between states and reduces attainable clock frequencies) and those need 100-300mV of overhead to do their job, so a bunch of voltages are now nominally ~200mV higher than they used to be to compensate and also as a buffer for the VRM's transient response.

1.5V is a common idle/light-load voltage because it gives the LDOs the most headroom for peak momentary boost.

I know that overclocking has and always will be at your own risk. At least with previous generations we knew what the max Vcore was, which made overclocking a bit easier.

 

[Redneck5439]

That's awesome on the 4.4 all core overclock, I am barley able to hit over 4.3. I have been running it pretty sustained at 1.387 and have yet to see anything negative voltage related.

Don't get me wrong, I'm happy with my rig. It is what I would consider above average for overclocking. Most of the 3800X systems I have worked on (only worked on 4, including my own) can hit 4.4Ghz all core @ 1.325 - 1.36V however seem to hit a wall thereafter. With my proessor I can also use CCX overclocking to set my fastest CCX to 4.25Ghz so half the cores are at 4.4Ghz and half are at 4.425Ghz @ 1.375V.

That's not bad, but I worked on a "golden" chip that I was able to get 4.45Ghz @ 1.36V. Unfortunately it wasn't mine and I couldn't keep it... I can't hit 4.45Ghz at 1.36V, but I can at 1.39V, however I just don't feel comfortable pushing that much voltage blindly. Temp wise I have a lot of headroom, but voltage wise is the big question.

 

[rigg42]

I have seen 1.325V thrown around, but with no technical data behind it
What do the experts out there have to say about 7nm Ryzen 3000 for sustained Vcore under load?

"According to FIT, the safe voltage levels for the silicon are around 1.325V in high-current loads
and up to 1.47V in low-current loads (i.e ST), depending on the silicon characteristics."

I'd say the FIT qualifies as technical data. It's probably unwise to go over 1.325 with an all core/per CCX overclock. I don't plan to go over that when I get my custom loop built. It's not currently an issue for me because I can't keep my 3900x cool at more than 1.2v in a high current load.

 

[InvalidError]

I know that overclocking has and always will be at your own risk. At least with previous generations we knew what the max Vcore was, which made overclocking a bit easier.

Since the on-chip voltage regulators have the final say on what the real Vcore actually is, AMD may simply have chosen not to publish a spec for an internally regulated parameter.

 

[Redneck5439]

I've done some more testing and it may be my motherboard, I'm using a Strix X470-F, so its not X570. I have noticed that my CPU under all auto settings with PBO enabled uses more power than my all core overclock of 4.4Ghz. With my multiplier overclock I'm running 4.4Ghz all core and my voltage under full load is a constant 1.373V. Using PBO I have managed to get an all core auto boost of ~4.3Ghz but the voltage as reported by Ryzen Master and any other monitoring software I utilize is a constant 1.41V under full load. That 1.41V isn't just the boost clock's normal behavior its the core voltage under full load. By that logic it would seem that my all core overclock is actually using less power under full load conditions than PBO.

I'm going to keep optimizing my PBO settings, and may even try a negative voltage offset to get the processor down to 1.373V all core under PBO. There is a small chance that would allow the overall all core boost to exceed 4.3Ghz. People running better X470 boards or X570 boards with better VRMs may not have the voltage usage enigmas I have seen with my system.