Can I Benchmark like this ???

Started by sushobhit, February 27, 2014, 07:21:14 AM

Previous topic - Next topic

sushobhit

Hi Friends ,

I was going through the advanced example and in the DrawFrame function there is a small code to calculate the FPS\
Something like : while gettimemillis < 1000 ; fps++ (gives the current fps)
Now , I want to know is that I want to use this code to Benchmark two phones (Graphics Processor)
Suppose one phone I have has some NVIDIA and the other has some other lesser like SHARED GRAPHICS PROCESSOR that shares the processing from chipset.
So will the FPS on the first one be greater than the second one.
Also will the difference be constant.
I want to do something like GET the FPS from the best GRAPHICS PROCESSOR (suppose I get 350 fps).. this value I save in my benchmarking program as the best value (the scale is of 10) so this value i.e 350fps is value 10 on scale.

Now the person runs Benchmark on his phone . My app (which is performing heavy graphics routine) calculates the FPS on the phone and compares it with the best i.e 350fps and accordingly gives the result to the user as to the quality of his GRAPHICS PROCESSOR.

The output would be something like this :
-----------------------------------------------------------------------------------------
Your FPS : 125                                                 Best FPS : 350
Your GC : Shared                                             Best GC : NVIDIA

           Benchmarking ...
                   
           Your score is 3.567   which is AVERAGE
           Consider upgrading your TAB/PHONE
           if you want to play games with high
           graphics....
------------------------------------------------------------------------------------


EgonOlsen

Just go for it...i fail to see the actual question in this post!?

sushobhit

I want to know if this will work ??

Is it practical ??

Is there a difference between FPS between different GC ??

And is it constant depending on their processing POWER ??

EgonOlsen

Yes, its ok to do it that way. Personally, i would render the benchmark scene like a game, i.e. add some code that decouples the game logic from the framerate and just count the total number of frames for a particular time frame. It did this for example here: https://play.google.com/store/apps/details?id=com.threed.jpct.benchxl