Are you checking if PBO exceeds the specified maximum clock speed?
Are you checking if PBO exceeds the specified maximum clock speed?
I performed thermal tests using Prime95 and Cinebench, but stability during lighter tests wasn't observed. A software update led to a system hang. Until the cooler issue is resolved, I've switched back to the stock version. By the way, Ryzen Master indicates that cores 3 and 4 are the most popular. After fixing the cooler, I'll think about trying an overclock again.
Each chip and motherboard behaves uniquely.
My 5600x and Prime x570 work with PBO in manual mode.
PPT 142, TDC95, EDC 140 – curve optimizer at +28.
LLC boost is high; all other core boosts and voltage settings are disabled. They consume excessive power.
Operating at 4.65 GHz for 3 or 4 cores based on load, with voltages between 1.21 to 1.26V.
At 4.5 GHz with all cores boosted, voltage sits between 1.28 and 1.32V.
For 6 cores, it’s 4.65 GHz with 1.36 to 1.41V for all core boost, which is very hot.
Every silicone component varies, and each core has its own characteristics.
The optimal core boost and voltage depend on the weakest core and its required voltage to achieve the desired speed.
My performance is reaching 4.9 with 1.4V for the single core enhancement (or 2 core boost). All cores operate at a reduced voltage, and the clock speed remains inconsistent. Prime95 shows lower results compared to other benchmarks like Cinebench.
The inconsistent clock speed is quite normal because the algorithm modifies individual core clocks up to 100 times per second depending on workload and temperature, along with other factors. Clocks also fluctuate constantly, typically lowering slightly during intensive all-core tasks like Cinebench. A very demanding task such as Prime95 small FFTs forces the algorithm to push clocks even further to maintain safe core temperatures. The main benefit of using PBO and Curve Optimizer with Ryzen is that the boost algorithm continues to function effectively, protecting the CPU from overheating during high loads—this is what you observe.
This constant adjustment makes clock monitoring largely meaningless when trying to assess performance. It only provides value if you run a timed benchmark throughout the test. Another benefit is that improved cooling occurs naturally because the algorithm doesn’t need to reduce clocks as much when cores aren’t overheating.
When applying Curve Optimizer, achieving optimal overall performance often involves lowering the EDC below stock settings—for example, reducing it from 142A to 120A for a 5800X. This makes the CPU less aggressive in showing many cores at maximum speeds during light threaded boosting, yet Cinebench BM scores remain higher. It seems this approach slightly reduces early core heating by trading some initial high clocks for a more consistent average over time.