Clarification needed for PBO (Watt, Temp, CPU Speed)

[Fireblade]

Hello Everyone,

Apologizes if this has been asked before but I did try to look and couldn't find any answers and if what I am saying is not clear enough but I am kind of new, actually returning after a really long break to gaming and computers.

My question relates to the PBO settings and the relationship between Watt, Temp and CPU Speed.

I have no need to overclock at the moment. What I am looking is the sweet spot in power consumption, performance and longevity. With that being said.

What I am really asking is do I keep lowering the Watt and Temp till I see my Speed being reduced and is there any point keeping the wattage and temp high? or there are other factors that I need to consider?

Currently I have a power limit of 75w and a temp of 70c as you can see from the screenshot but the speed is not reduced compared to my previous settings which was 85w and 85c temp and still was getting a 5,500Mhz speed.

So is there a point having higher values on Watt and Temp?

Also in HWiNFO I cant seem to find the SVI2 setting anyone knows what's up with that.

Let me know if you need any other info or readings on my part.

Thank you !

 

[Lost_Troll]

Currently I have a power limit of 75w and a temp of 70c as you can see from the screenshot but the speed is not reduced compared to my previous settings which was 85w and 85c temp and still was getting a 5,500Mhz speed.

When you set a lower PPT and temp in PBO, you will sometimes see the max boost clock for the CPU, but your average running clock speeds will be lower.

To get higher average clock speeds and lower temps, you can do a manual under volt with PBO set to a negative curve of say -20. This will lower the temps and power while giving you higher average clock speeds.

AMD's Understanding Precision Boost Overdrive in Three Easy Steps

There are some good videos on how to setup PBO with and without the manual under Volt.

Easy 7800X3D Tuning and PBO2 Undervolting Guide

Kicking Ryzen 7000 into HYPERDRIVE! Overclocking, Curve Optimizer, 5.85 GHz!

How to Overclock / Undervolt Ryzen 7000 & 8000 CPUs | 2024 Tutorial

 

[Fireblade]

I forgot to mention in my initial post that I do have Negative 30 in all cores.
But after what you said about Max. clock speed, I ran two different tests watching live the clock speeds.

Test 1: Negative 30, 80c limit, Auto setting for power

This got a 1140 Cinebench score with Temps as high as 82c and watt running up to 115w and Vcore to 1.424V....
The Live clock speed were running around 5.300Mhz

Test 2: Negative 30, 70c limit, 85w limit

This got a 1090 Cinebench score with Temps as high as 65c and watt running up to 87w and Vcore to 1.304V....
The Live clock speed were running around 5.100Mhz

Are those Values good with curve optimizer I mean the Voltage does seem quite high. Can someone check the Screenshots and let me know if there is a way to maintain high clock speeds with lower temps and watts? Thanks again!

 

[Zach_01]

First of all you have hidden a lot of HWiNFO sensors.
For CPU voltage you need the SVI3 TFN sensor.
For power you need CPU PPT (W) and CPU PPT limit (%).

Second, you have to get clear about what you want from the CPU.
You have to distinct single thread speed and all core speed.
There are different actions to restrict single thread boost and all core boost.
By lowering power (PPT limit) you only restrict all core load and speed and not single core speed.
Ryzen 7000 CPUs are sometimes hitting max temp with single core speed and some others with all core speed.

For a Ryzen CPU up to 1.45~1.5V under single thread load or low load is very common and normal.
Only under full load voltage drops under 1.4V and as the load gets heavier and heavier (different type) you might see it all the way down to 1.0~1.1V.
As the load gets progressively higher the voltage drops progressively from 1.5 to 1.1V or even lower. Max speed is also progressively dropping.

 

[Fireblade]

Ευχαριστω φιλε.

Interesting. Thanks for the explanation about Voltage and single core Temp.

The sensors I have kept is because after researching in the HWiNFO forum are the most reliable ones. I did add your suggestion about "CPU PPT" to see the values and it is pretty similar with the "CPU Package Power" that I was running.

The "CPU PPT Limit" on the other hand I have no clue why we need it as it is a percentage.

And about the Voltage yes I added the "CPU VDDCR_SOC (SVI3 TFN)" since I cant find any other close to the one you mentioned about CPU voltage?

After what you said I am thinking that basically if you put negative 20-30 and just a temp limit it also adjusts the power limit automatically? Is that how it works?

What I seem to don't understand is how you can feed the CPU with more power and lower the temps and still have it perform the same or better? Because that is what is basically happening? Anyway I am confused now...

P.S I dont think HWiNFO has an average for CPU clock.

 

[Zach_01]

The SOC SVI3 TFN is the I/O die voltage, not the CPU core voltage.
You need also the VDD_SVI3_TFN.
That’s the Vcore.
And those sensors are read straight from the CPU, like PPT/EDC/TDC also. The Vcore and package power is from board sensors that are less accurate.

The PPT limit (%) is just an indication of the power limit. Not very important but useful for others trying to read your screenshot. If I see PPT at 80W and PPT limit at 80% I will instantly know that your limit is 100W. Same for EDC. That’s the current (A) going through the entire CPU package.

Also you might want to enable “Snapshot CPU polling” from main settings of HWiNFO for even more accurate readings.

I will come back later as I’m still at work and don’t have much free time now.

EDIT:

After what you said I am thinking that basically if you put negative 20-30 and just a temp limit it also adjusts the power limit automatically? Is that how it works?

Not exactly...
Lets say you have a power limit of 100W and a temp limit of 80C.
The CPU on auto boost and under any load will always try to maximize the available power so you can have the best performance, unless you have set the system (windows) to power saving mode.
It will also try to reach max core speed under single/low-medium loads.
When you restrict the temperature it will boost until it hit that thermal limit. What will be the power at the point of thermal limit its no straight answer.
It will depend on a few variables like the cooling capacity, the type of load, how many cores/threads are loaded, what's the current through the CPU (Amperage / EDC sensor) and maybe more...

When you set curve optimizer to neg steps you are telling the CPU to boost at whatever speed it can achieve (respecting power and temp limit) but with less voltage.
On quick thought this seems like lowering total power but that is not the case. Because of the undervolt (through V/F curve remapping) the CPU now has room to boost more.
It will always try to maximize power and temp limit if all other limits are respected, like EDC for example or the max boost speed.
So curve optimizer on neg causes the CPU to boost more within the same power and temp limit. So gain performance on the same PPT, temp, EDC.

If someone set a really low temp limit like 65~70C for example then most likely the CPU will never reach the power limit. For single thread boost it will probably still hit max speed but it will be even shorter bursts from the default boost when temp limit is 90~95C.
This will be shown directly on the average core clock column and the also right below on the average effective clock column.
If cooling capacity gets higher, like increasing fan/pump speeds, lower ambient/room temp or even installing a bigger/better cooler then the boost will increase by a margin depending on how much more heat is removed per amount of time.

Personally I have a loose PPT limit (200+) and I only care about temperature and current (A). I've set a limit of 80C and a 180A EDC and I let the CPU to boost and draw power as high as it can.
Though I am not running too often all core loads. My regular usage is gaming or having many things running at the same time like browsing with 30~40tabs, watching videos, running VMs (emulators) all together.
Under all the above tasks the max CPU temp is usually around 75C with an average of 40~65C depending on what I'm doing at any given moment.

I am not editing or rendering anything that uses 100% of CPU. That would cause the CPU to go straight to 80C with a 200+W limit. If I set max power to 5900X default (142W PPT) then under all core load the max temp is usually around 65~70C. Still under single/low-medium thread load the max temp is around 75C.

Also my windows power plan is set to balanced and sometimes I switch CPU behavior through "secondary" settings.

This is always on balanced


and this I change depending on what I'm doing

 

[Fireblade]

Hmm interesting. Yes it makes much more sense the way to explained it now. I do have several questions but I will let the information sink in first and come back later.

I will say this for now though. The reason I wanted lower temps is because my CPU (since its been a while I didn't game) is going to get under some heavy based CPU gaming, games like POE which are really stressing the CPU with loads of calculations and it will be for many hours. So basically I wanted a lower temp and less power because it gave me anxiety to see the CPU operate at so high temps (75c+ I consider it high) for so many hours and having my computer sound like a nuclear reactor. But that maybe be just me because its been a while since I had a proper gaming PC.

 

[Zach_01]

Here is a couple of months old screenshot while gaming for about 2 hours. I've reset sensors and start the game. When I got out of the game I took the SS.
As you can see max temp and avg temp is 2 very different things.
And CPU cooling by avg is on low rpm during the game for 99% of time.