Ryzen automatically limiting power when temperatures goes beyond 73°

[Islam Ghunym]

The build is:

ASUS B450M-A

Ryzen 5 3400G (65W TDP)

2 x 8GB of micron Rev E memory

 

I started with overclocking memory to 3466 MHz and shrinked every timing as well (couldn't stabilise the 3533)

 

added a negative offset on Vcore (-0.043)

I enabled PBO X10 scalar +0MHz CPU (couldn't add even +25 without having instability)

+200MHz GPU.

Disabled PPT, EDC, TDC limits and platform thremal limits.

 

The build suffered from sudden reboots on 3D applications and stress test. To solve the problem a 1.243V SOC was required with PBO enabled it can go as high as 1.32V in some rare cases on 3D applications, but it is almost always around 1.25V so I am fine with that. The memory was stable at even 1.35V SOC so these high voltages are not a problem. (I am aware that unsafe voltages starts beyond 1.3V so going to 1.34-1.36V and staying there for extended time is going to cause a degradation and that voltages reported by softwares like HWiNFO are not exactly the real ones)

 

Now after running any game, the CPU package power goes (75-87W)

Later once temperature goes beyond 73° the package power automatically gets limited to 65W only. The difference in game FPS is not clear (almost the same), but the CPU clock speeds goes lower and graphics core clock shows the same behaviour. I don't like this limit. How can I disable this limit?

[Chiyawa]

Hmm... Well, one way to elevate the restriction is to get a better cooler. PBO will deactivate if the temperature rise above 70c, though I'm not sure about this, but my Ryzen 3600 began to reduce the CPU clock little by little from all core 4100MHz to around 3900MHz while maintaining around 71c.

 

I use Cooler Master Hyper 212 LED Turbo cooler for my CPU, and it took about more than 30 minutes of stress test to drop the frequency from 4.1GHz to 3.9GHz all cores. Maybe a better cooler can help?

[Islam Ghunym]

4 hours ago, Chiyawa said:

Hmm... Well, one way to elevate the restriction is to get a better cooler. PBO will deactivate if the temperature rise above 70c, though I'm not sure about this, but my Ryzen 3600 began to reduce the CPU clock little by little from all core 4100MHz to around 3900MHz while maintaining around 71c.

 

I use Cooler Master Hyper 212 LED Turbo cooler for my CPU, and it took about more than 30 minutes of stress test to drop the frequency from 4.1GHz to 3.9GHz all cores. Maybe a better cooler can help?

That can be a solution, but from my understanding to PBO scalar is that it will allows the CPU to stay boosted at a certain speed by increasing the voltage when the temperature is higher than where it should be stable so:

If it was originally doing 3.9 GHz at 1.45v when temperature is 80 and then drops to 3.85 when temperature goes beyond 80° to 85° for example. Here PBO is going to keep allowing 3.9 GHz even beyond 80° by giving more voltage 1.47v which will also increase the temperature a little again, but not much to effect the advantage of higher Vcore.

 

The thing is: if PBO is combined with PL, it will end up with more like no PBO.

Not only that, but PBO also controls power limits which by default is 88w so it should allow more than 88w, but what is happening is that I am even ending up at 65w only and everything I did to OC was meaningless except the memory OC.

Manual overclocking ends up with 3950MHz all the time even at 95° with graphics clock at 1550 MHz. The performance I got was almost the same I got with PBO and current power limits situation because of high temperatures limiting maximum OC I can go with. PBO could do about 40.75 GHz while running a game which is better than limiting my CPU to 3.95 only and could boost the graphic to 1480 MHz only. I got almost the same FPS.

If just PBO works up to 95° without limits, I will definitely get much better performance.

[Chiyawa]

12 hours ago, Islam Ghunym said:

*Snip

I see. So far I don't overclock myself, everything is set to auto. This is the behaviour I observed during stress test.

 

I wonder if you can use Ryzen Master to adjust some of the setting as well.

 

Anyway, I wonder if your mother board couldn't keep that power up with your APU. According to the tier list from this thread, it seems that this board VRM supply a maximum of 50A current draw. Check if your VRM is overheating or not. I think this is most likely your VRM couldn't keep up with your overclock APU.

[Jurrunio]

The problem is the SOC VRM, the one that powers the iGPU. Your board is too underbuilt for overclocking tbh, get a powerful fan and blow at the VRM directly.

[Islam Ghunym]

On 5/25/2022 at 5:03 AM, Chiyawa said:

I wonder if you can use Ryzen Master to adjust some of the setting as well.

 

Anyway, I wonder if your mother board couldn't keep that power up with your APU. According to the tier list from this thread, it seems that this board VRM supply a maximum of 50A current draw. Check if your VRM is overheating or not. I think this is most likely your VRM couldn't keep up with your overclock APU.

Probably too late answer, but no using Ryzen master does not do anything different. However on old AGEISA builds on the same motherboards, the CPU had no limits and was pulling as much as 120-185 watts without limits. The VRM was also bellow 90° on the multimeter all the time during hot summer environment. I heard about this motherboard having bad VRM and that it is not designed for overclocking. I see also that the VRM can do 2 x 25A only for the SOC, but it was giving about 120W just on the SOC at 1.19v. so it comes down to what being able to do 50A mean. I see that this is maybe the guaranteed amount of power the VRM can deliver for the SOC with Extreme LLC without having problems with the output voltage like not being able to deliver the required amount of voltage causing instability. I noticed that the motherboard can't do LLC at all. If I increase the LLC level to medium the VRM won't be able to deliver the required voltage causing system to reboot. However using PBO does not initiate that rebooting behaviour. PBO is a much better way than using LLC as I experienced. At the very recent AGEISA firmwares the power draw has been locked to 88W for the APU and even initiate a limit at 65W once temperature appeared to be higher than 73°.

 

What I am trying to say is that you are right about this motherboard VRM being bad, but the limit was because of AMD's recent limitations and what matters now is that I will be fine without these limits because I was getting higher performance and VRM temps were fine. Theoritical information does not always describe every situation in reality.