Ryzen reduces performance when heat exceeds 73 degrees.
Ryzen reduces performance when heat exceeds 73 degrees.
The system uses an ASUS B450M-A with a Ryzen 5 3400G processor. It has 8GB of memory in a micron Rev E package. I initially boosted memory speed to 3466 MHz and adjusted timing parameters, but couldn't get the voltage to stabilize at 3533MHz. I added a negative offset on the Vcore and enabled PBO X10 with a CPU frequency of 0MHz and a GPU boost of 200MHz. I turned off PPT, EDC, TDC limits and platform thermal caps. The build experienced frequent reboots during 3D apps and stress tests. To fix this, a 1.243V SOC was needed with PBO enabled; in rare cases it reached up to 1.32V for 3D apps, but usually stays around 1.25V. Memory remained stable even at 1.35V SOC, so those voltages aren't critical for me. High voltages above 1.3V can cause long-term damage, and software readings may not reflect real performance. After gaming, the CPU draws 75-87W, and once temperature exceeds 73°C it caps at 65W. Game performance drops slightly, but CPU speeds and GPU clocks behave similarly. I’m uncomfortable with these limits. How can I remove this restriction?
Consider upgrading to a more efficient cooler. PBO shuts down when temperatures exceed 70°C, though I'm uncertain about this. My Ryzen 3600 gradually lowered its CPU clock from 4100MHz to about 3900MHz while keeping temperatures near 71°C. I've been using a Cooler Master Hyper 212 LED Turbo cooler, and it took over 30 minutes of stress testing to bring the frequency down from 4.1GHz to 3.9GHz across all cores. A better cooler might make a difference?
This approach might work, though my knowledge suggests PBO scalar enables the CPU to maintain higher speeds by raising voltage when temperatures rise above optimal levels. For instance, if the chip runs at 3.9 GHz at 1.45V under normal conditions, it can sustain 3.9 GHz even when temperature climbs past 80°C to around 85°C, thanks to increased voltage to 1.47V which slightly raises the heat again but preserves the higher Vcore benefit. Combining PBO with PL would likely restore near-native performance without PBO. Additionally, PBO manages power caps—normally set at 88W—but high temps still cap output. Without PBO, I’m stuck at around 65W despite all other adjustments, making manual overclocking ineffective except for memory tuning. Running at 95°C with a 1550MHz clock still gives me solid performance, around 3950MHz, matching what I’d expect with PBO enabled. If PBO alone could run at 95°C, I’d see a noticeable performance jump.
I understand. I haven’t increased my settings beyond the automatic mode, which is what I noticed during the stress test. Would it be possible to use Ryzen Master for further adjustments? Also, could your motherboard handle the additional power from your APU? Based on the tier list in this discussion, it looks like the board’s VRM can only supply up to 50A. Make sure your VRM isn’t overheating, as that might be the issue. It seems your VRM likely couldn’t keep up with the overclocked APU.
The issue lies with the SOC VRM, which supplies power to the iGPU. Your board isn't strong enough for overclocking, so consider adding a better fan and directly targeting the VRM.
Likely too soon to conclude, but not relying on Ryzen master changes anything much. On older AGEISA boards built with the same chipsets, the processor ran at full capacity drawing up to 120-185 watts without restriction. The power delivery unit stayed under 90° during hot summer conditions. I’ve heard stories about this board having weak VRM and not being built for overclocking. It seems the VRM could only handle 2 x 25A for the SOC, yet it was delivering around 120W just on that component at a voltage of 1.19v. So the real issue is whether it could safely supply 50A. I think this reflects the maximum power the VRM can provide for the SOC with Extreme LLC without risking voltage instability. I noticed the board doesn’t support LLC at all. Raising the LLC level to medium would cause the VRM to fail and trigger a reboot. But with PBO, that behavior doesn’t happen. PBO is a much improved alternative compared to LLC, which I personally found more reliable. Recent AGEISA updates have capped power usage at 88W for the APU and even limit it to 65W when temperatures exceed 73°C. What I’m saying is correct about the VRM limitations, but they stem from AMD’s recent restrictions. What really matters now is that I won’t face those caps and can still achieve better performance and stable temperatures. Theoretical data doesn’t always match real-world results.