Optirun and Bumblebee basic question

Question

I have a laptop which has two GPU's in it, an Intel and an nVidia. I have been using Bumblebee to switch between the GPU's, for instance running glxspheres or glxgears in a terminal the Intel GPU is used and the terminal shows me the frames per second, which is about 60FPS and I understand this to be normal because the refresh rate of the screen is 60Hz. If I run optirun glxspheres or optirun glxgears bumblebee switches to using the nVidia GPU but the FPS remains at 60FPS because it is limited by the screens refresh rate.

So my question is this, if the FPS is limited by the screens refresh rate and the Intel GPU can perform at that FPS, what is the advantage of using the nVidia GPU at all?

Answer 1

You seem to have rushed to a conclusion that the Nvidia has no benefit at all, but it's based on at least two severe misunderstandings about what that (toy, non-) test described tells, and they could've been avoided by searching first!

(A) Testing with vsync enabled and a trivial non-benchmark like glxgears will make any 2 barely modern GPUs 'seem equal', if your definition of "equal" is merely 'can both push out 60 FPS using an extremely simplistic rendering'.

(B) You could at least disable vsync, to remove that artificial ceiling. See Disable vertical sync for glxgears

(C) glxgears may even unfairly punish the discrete GPU due to PCIe bus saturation. You're barely taxing the real machinery of the GPU at all, just making it render toy frames at high rates and push them over PCIe for the integrated GPU to send out to the screen. You need to benchmark the discrete GPU using something that gives it real work to do - crunching polygons and such, not just spamming frames. See this explanation at the Bumblebee tracker and why is glxgears faster without optirun and etc.