Question AMD Ryzen 5600x PBO speed dips

A Gamer

Honorable
Mar 29, 2016
52
1
10,535
0
My score isn't all too bad, but i wonder why my CPU Freq. goes back to lower speed sometimes.
Here is a picture: View: https://imgur.com/fAxcuew

Is there something i can do about this? (No manual settings)
Score:
GPU: 15059
CPU: 8254
13401
Temps: CPU Max. 81C

Settings:
PBO: ON
PBO Limits: Motherboard
Curve: Negative: -25 -25 -15 -25 -15 -25
that's all i changed + asus performance setting to keep the clocks higher.
 
As there is a score, is this some type of benchmark?

If so, you cannot assume the load is100% consistent for every second during the test, as there may be brief transitions from one portion of benchmarking to another causing the load to decrease, with a brief commensurate drop in clock speed. (A cpu does not/should not just sit at a clock speed 24/7 under no load for no reason at all, and, outside of benchmarking, such as routine desktop operations, clock speeds across all cores can and do fluctuate many times per second from as low as 800 Mhz to as high as 4.5 GHz or so, or, higher on other Ryzen 5800X, 5900X models...

If you want to see if your CPU can maintain clock speed under load, then run HWMontitor to monitor assorted core clock speeds, and, induce a constant load with CPU-Z/bench/'stress cpu'....(be sure to hit 'stop' when thru observing to remove the induced load to avoid wasting power/100% CPU usage.)
 
Last edited:
My score isn't all too bad, but i wonder why my CPU Freq. goes back to lower speed sometimes.
Here is a picture: View: https://imgur.com/fAxcuew

Is there something i can do about this? (No manual settings)
Score:
GPU: 15059
CPU: 8254
13401
Temps: CPU Max. 81C

Settings:
PBO: ON
PBO Limits: Motherboard
Curve: Negative: -25 -25 -15 -25 -15 -25
that's all i changed + asus performance setting to keep the clocks higher.
that looks like a 3DMark monitoring chart after a GPU benchmark test.

Since it IS a GPU benchmark it's dependent on GPU performance far more than CPU, at least until the CPU performance portion. So if CPU frequency drops in clocks during the GPU rendering tests it's because the CPU boosting algorithm sees a moment to take a break from holding high clocks.

Even though the graph makes it look like a continuous reading, it's not. What you're seeing in the chart is some sort of a smoothed average of all the cores' clocks for some unknown number of polls. Between each polling the CPU has several million (billions?) cycles for each core where it's spending a LOT of time at a lower clock...and some time at a higher clock too. The monitor program polls (asks) the CPU every half second or so what each core's clock speeds are and then plots it.

The smoothing algorithm 3DMark uses to make the chart is unknown but it's obviously trying to show as a smoothly leveled clock speed what are actually very dynamicaly changing clocks for all the cores of the CPU.
 
Last edited:

A Gamer

Honorable
Mar 29, 2016
52
1
10,535
0
As there is a score, is this some type of benchmark?

If so, you cannot assume the load is100% consistent for every second during the test, as there may be brief transitions from one portion of benchmarking to another causing the load to decrease, with a brief commensurate drop in clock speed. (A cpu does not/should not just sit at a clock speed 24/7 under no load for no reason at all, and, outside of benchmarking, such as routine desktop operations, clock speeds across all cores can and do fluctuate many times per second from as low as 800 Mhz to as high as 4.5 GHz or so, or, higher on other Ryzen 5800X, 5900X models...

If you want to see if your CPU can maintain clock speed under load, then run HWMontitor to monitor assorted core clock speeds, and, induce a constant load with CPU-Z/bench/'stress cpu'....(be sure to hit 'stop' when thru observing to remove the induced load to avoid wasting power/100% CPU usage.)
This is 3DMark, Time Spy. The blacklines indicate a switch of scene/test. The dips are not always on the same place and not always in the same amount.
 
I'm seeing a similar pattern in my testing, where the CPU speed fluctuates between base and boosted speed. Except when it gets to the CPU test where it caps out at the boosted speed. I've also seen this behavior in Fire Strike where the CPU speed dips during pure graphics tests but not during tests where physics are used.

My guess is that during the graphics tests, the GPU is taking long enough that the CPU has found moments where it can clock down because it can't send anything to the GPU (because it's busy). For comparison, the other tests like Sky Diver, Cloud Gate, and Ice Storm, all keep the CPU speed boosted, but at the same time I'm also getting like 200+ FPS in all those tests.
 

ASK THE COMMUNITY

TRENDING THREADS