FPS performance measuring

[citizen6]

I find the way that tech channels in Youtube assess performance of GPU performance in games rather unscientific and unreliable.

 

I believe tech channels should use additional (better) tools and metrics for analyzing their performance.

 

The mean FPS by itself is really not a reliable metric. Including metrics such as variance and sampling rate could be a start.

 

Consistency in performance is really important and ignoring this could create false perceptions.

 

For me, this  kind of performance testing could stand out of the crowd. 

 

[Mira Yurizaki]

A lot of major review websites, like Tom's Hardware and Anandtech, are doing this. FRAPS also can log frame times.

 

However the other stick in the mud is that frame rate variance is mostly a CPU dependent metric. So if you're going to pair a GTX 1080 with a Core i3, you bet your butt you'll get more frame rate variance than if you paired it with a Core i7.

[rf9661]

20 minutes ago, citizen6 said:

I find the way that tech channels in Youtube assess performance of GPU performance in games rather unscientific and unreliable.

 

I believe tech channels should use additional (better) tools and metrics for analyzing their performance.

 

The mean FPS by itself is really not a reliable metric. Including metrics such as variance and sampling rate could be a start.

 

Consistency in performance is really important and ignoring this could create false perceptions.

 

For me, this  kind of performance testing could stand out of the crowd. 

 

I think that a lot of the bigger tech YouTube channels do this since it's what a lot of amateurs want to know. More in depth people (like us) want to see the pure statsitics, unsugarcoated.

[citizen6]

Quote

However the other stick in the mud is that frame rate variance is mostly a CPU dependent metric. So if you're going to pair a GTX 1080 with a Core i3, you bet your butt you'll get more frame rate variance than if you paired it with a Core i7.

 

I agree it is hardware depended but most of the time, at least on LTT, they use the same hardware bench to do their benchmarks. So this metric does make sense. But I do not know if it will make a difference in different hardware. I believe that if the GPU is close to its saturation point and fps are low, the variance will increase and this will reveal info about the GPU architecture efficiency.

[citizen6]

Quote

I think that a lot of the bigger tech YouTube channels do this since it's what a lot of amateurs want to know. More in depth people (like us) want to see the pure statsitics, unsugarcoated.

More informed consumers will motivate to make better products for all of us. People who dedicate significant amount of their day in tech products maybe they would like to learn something more. I might be wrong though.