erkexzcx
commented
3 months ago Hi. Thank you!
can you please explain what the FPS stabiility graphs
Specifically there are 2 sections - one for FPS (which is a "final product") and another one for Frametime (which is "root cause" for FPS values). For example, look at this FPS chart of yours uploaded benchmark:
You can clearly see that FPS was way more stable on green line, while blue line was always spiking up and down. Mathematically, there is a way to calculate a "stability score" which is basically a "Standard Deviation": https://en.wikipedia.org/wiki/Standard_deviation
Here is the relevant part from wiki:
So in short - click on these buttons to disable/enable certain metric:
And then you have these values.
Let me try to explain again, but differently. Look at this chart:
Now, you have a feeling that CachyOS has much better stability, less FPS stutters and so on. Your friend, however, says the opposite, that Nobara actually gives better stability. Now, this argument might lead to ER (emergency room) visit or permanent damage to your brain tissue if your friend is strong enough. However, a civilized approach to this argument is to use mathematical approach to calculate the actual FPS stability "score":
According to this data, CachyOS gives better FPS stability (less spiking in terms of FPS) in comparison to NobaraOS. Congratulations, without a physical fight you proved your theory to your friend that CachyOS has greater FPS stability in comparison to Nobara OS. :sweat_smile:
Disclaimer - in order to truly compare FPS stability, you should collect each frame instead of frame at certain intervals (e.g. 100ms). In MangoHud, interval should be set to 0, so it captures each frame. I am not sure if the same is possible in Afterburner (never tried).
This is because when using e.g. interval of 500ms, MangoHud might skip stutters from time to time, so final recorded benchmark data would not represent a reality.
These FPS stability charts were added by me when I was participating in sched-ext discussions and to prove my point that bpfland gives superior FPS stability on Steam Deck playing Cyberpunk 2077 where all other schedulers gave me stutters all over the place and made the game unplayable. On the other hand, bpfland game me "dynamic FPS lock" where there are no stutters, but game fails to reach 60 fps and constantly changing.
thesacredsin
Hi, thanks for the great tool. Just a question can you please explain what the FPS stabiility graphs represent, and is possible to maybe add a % more/less stable metric, in other words Lets say A's FPS is 5% more stable than B. At the moment, to me atleast the metric does not say much.