Graphics Cards

[m1hhe]

Hello!

PC:
Ryzen 5 3350G - cooling DEEPCOOL GAMMAXX 200T
MSI B450M-A PRO MAX
2x4GB RAM - Crucial CT4G4DFS824A

Faced such a problem that in games it starts to lag: after a minute, FPS drops by 15 seconds, and then again it is normal for 1 minute.
The RAM was overclocked to 3200MGz, thinking that it was in it, dropped the overclocking to 2400MGz, but the problem remained, although lags became less frequent - every 2 minutes lags for 15 seconds.
When I installed a discrete video card Nvidia GeForce GTX 650 Ti, no lags occur at any available RAM frequency, but the built-in Vega 11 is certainly better than my video card, I would like to play on it.
In addition to games, the problem is also observed in FurMark, and then I realized that the matter was in the integrated video card.
I must say right away that the processor is without overclocking, as is the frequency of the integrated graphics core.
Here is a link to YouTube, where I recorded the problem in FurMark, where you can see how the frequencies of the integrated graphics drop after a while. The problem can be seen at 2:30: https://youtu.be/YfveNR_zyso

[alseu]

Hello
Thanks for the video
Can you perform the same test with HWinfo64 opened to monitor power usage and voltages during the test to ensure that the frequency drop doesn't occur due to throttling (like it happened here https://community.amd.com/t5/processors/ryzen-5-3400g-cooled-very-well-but-it-s-thermal-throttling-f...)
Is your motherboard B450M A-Pro Max or B450 A-Pro Max?

[m1hhe]

Thanks for the answer!
My motherboard is MSI B450M-A PRO MAX.
Here is a video from HWiNFO64, it seems to have brought everything that is needed.
Lags at 2:50

https://youtu.be/oaEs06zwhIo

[alseu]

Good video!
As you can see hwinfo64 shows that throttling occurred due to  POCHOT EXT assertion which is caused by some external component overheat on your motherboard.
SoC power consumption is 45-46W in this test which is too much for the SoC VRM on B450M A-Pro without additional direct cooling since it has no radiator and budget components.
VRMs reach 105-120C and throttle (HWinfo sensor shows 118-119C but that doesn't seem right since it was the same at the beginning of the video, when there was very low power load, looks like a "stuck" sensor value). This is in spec but usually NOT recommended for the VRMs and might be dangerous for the capacitors nearby - their max operating temperature on this motherboard is 85C.
Have to either limit fps in mentioned games/ set direct airflow over these VRMs/ get another motherboard with SoC VRM heatsink/ set negative SOC voltage offset in BIOS (requires tuning and might produce instability)



Good luck!

[m1hhe]

Thanks again for your answer.
How can I solve this problem programmatically or how would you advise to cool the VRM?
And I would like to know why I had this problem if I saw on YouTube how the same processor was installed in a cheaper motherboard and even overclocked, and everything worked.

[alseu · Accepted solution]

Software solution would be either to reduce framerate in games you play (ex: Rivatuner) or (and) to downvolt the SoC by setting negative voltage offset (ex: -0.08V) or constant voltage (ex: 1.02V) in BIOS.
To improve SoC VRM cooling you can use a classic cooler instead of tower one (ex: ryzen box cooler), to add a cooler on top of the case if it has that slot, to add a direct airflow over VRM (ex: picture), to buy and install a radiator with thermal pads (aliexpress market). Depends on your budget, pc case and skills.
I can't comment the question about YT content without the video. There are a lot of variables.