Isaac Sim Version
4.0.0
Operating System
Ubuntu 22.04
GPU Information
- Model: Quadro RTX 4000
- Driver Version: 535.183.01
- CUDA Version: 12.2
Topic Description
Hello I’m using the python interface to isaacsim to view some simulation data. Regularly I get errors on startup like:
2025-05-08 13:52:11 [1,011ms] [Error] [carb.cudainterop.plugin] CUDA error 999: cudaErrorUnknown - unknown error)
2025-05-08 13:52:11 [1,011ms] [Error] [carb.cudainterop.plugin] Failed to query CUDA device count.
2025-05-08 13:52:11 [1,011ms] [Error] [carb.cudainterop.plugin] Could not query CUDA device.
2025-05-08 13:52:11 [1,011ms] [Warning] [carb.cudainterop.plugin] Unable to get NVML device from physical device.
There is then continous spew: [Warning] [omni.hydra] Rendering failed.
Often this seems to happen after my machine has been suspended. The only way that I can fix is to reboot my machine. Oddly enough if I use the omniverse-launcher to start omniverse code (2022.3.3) it seems to render fine even as isaac is hosed.
Error Messages
2025-05-08 13:51:18 [32,790ms] [Warning] [omni.graph.core.plugin] OmniGraphSettings::getCudaDeviceOrdinal: unable to get a valid CUDA device id from the renderer. Defaulting to GPU0.
2025-05-08 13:51:18 [32,795ms] [Error] [gpu.foundation.plugin] CUDA interop cannot be used!
2025-05-08 13:51:18 [32,795ms] [Error] [gpu.foundation.plugin] Texture creation failed for the device: 0.
2025-05-08 13:51:18 [32,795ms] [Error] [carb.scenerenderer-rtx.plugin] Failed to allocate 1280x720 LdrColor resource for device mask 0x1
2025-05-08 13:51:18 [32,795ms] [Warning] [omni.hydra] Rendering failed.
Additional Information
What I’ve Tried
Rebooting, a/b testing isaacsim vs omniverse code
Would love to have a way to reset the state of things so I didn’t need to restart machine, which is making me nervous about deploying things to CI.