AMD kicking Nvidia's ass in OpenGL?

Get your attention? hehe… I thought it might.

I’m a huge fan of Janus VR and the problem currently is that we can get much better framerates with an AMD card than we can in any Nvidia cards. For comparison, I can get Janus VR to run at around 200 FPS with a Radeon HD 7770 but with my Nvidia GTX 980 I’m getting around 60 max (yes vsync off).

The core of the problem seems to be the OpenGL calls that Janus VR is making. So the question is:

Is there a list of known OpenGL calls that are considerable slower on Nvidia architecture than AMD so that we can try to optimize and get users with Nvidia cards better performance?

As the last post in the linked thread is suggesting, try getting away from legacy APIs like immediate mode and display lists which are not available in newer OpenGL core versions anymore and port to VBOs.

Immediate mode and display lists are highly optimized for workstation applications because they still use a much wider range of OpenGL API entrypints. The consumer drivers are tuned for game like content which are normally using available modern APIs earlier due to much shorter longevity of the software.

Shameless plug: If you need an OpenGL renderer backend which really shines with respect to triangle throughput and parameter update performance while using modern OpenGL features you might want to have a look at the “RiX” module in our open source nvpro-pipeline.
More details here:

Thank you so much for the reply Detlef! Great info and I’ll replay that.