[Question] GUI rendering not in real-time
Hi all,
I'm running the play script in my local computer (bare-metal) with cpu (Intel i-9 11th gen).
I'm trying to run only 1 humanoid robot with sim_dt=0.001 and decimation=10
[INFO]: Base environment:
Environment device : cpu
Physics step-size : 0.001
Rendering step-size : 0.001
Environment step-size : 0.01
Physics GPU pipeline : False
Physics GPU simulation: False
But I'm experiencing slow motion on my GUI rendering.
From my past experience with IsaacGym, if I run the play script with cpu device, the GUI rendering was real-time speed. (using cuda:0 is slower than using cpu)
However, now in Orbit, even with the cpu device, the rendering speed seems almost x 3 ~ x 5 times slower.
Of course if I increase sim_dt and lower decimation (e.g. sim_dt=0.005, decimation=4), then the rendering gets faster.
I want to keep my original params (sim_dt=0.001 and decimation=10) while rendering in real-time speed.
Is there a way that I can get the real-time speed? Or is it just bottleneck by the speed of the physx and renderer?
Thank you!
I'm interested on this too, I'm using the same params and experiencing a slow motion simulation
I used the latest IsaacLab, and found that the self.sim.render() takes too long (5 seconds). Any comments?
Isaac Lab rendering performance could also depend on the GPU hardware available on the machine. The renderer in Isaac Gym was quite simple and didn't have ray tracing capabilities, while the renderer in Isaac Lab requires a lot more compute. You could try adjusting some of the render setting parameters from the UI to lower the rendering quality a bit, which could help with performance. Some settings we use for rendering could be found here.