ERROR_INCOMPATIBLE_DRIVER when using Isaac Lab?

Omniverse Isaac Sim
, , ,
1

Isaac Sim Version

5.0.0

Isaac Lab Version

2.2.1

Operating System

Other (please specify): CentOS 8.1 (cluster)

GPU Information

  • Model: TITAN RTX
  • Driver Version: 560.35.03

Topic Description

When RL training in Isaac Lab (using the default RSL-RL training code) on a TITAN RTX (driver version 560.35.03), I get the following errors:

2025-09-26T11:13:06Z [11,597ms] [Error] [carb.graphics-vulkan.plugin] VkResult: ERROR_INCOMPATIBLE_DRIVER
2025-09-26T11:13:06Z [11,598ms] [Error] [carb.graphics-vulkan.plugin] vkCreateInstance failed. Vulkan 1.1 is not supported, or your driver requires an update.

I see that the Isaac Sim technical requirements state that:

Versions greater than 535.255 have their version reported incorrectly in Vulkan, causing the driver to be misidentified as incompatible. A workaround is to disable the driver verification check with the command-line argument ā€˜ā€“/rtx/verifyDriverVersion/enabled=falseā€™

Iā€™ve tried applying this argument as follows:

./isaaclab.sh -p scripts/reinforcement_learning/rsl_rl/train.py --task=Ext-Isaac-Velocity-Rough-Anymal-C-v0 --headless --kit_args=ā€œā€“/rtx/verifyDriverVersion/enabled=falseā€

That didnā€™t work, so I then tried adding the argument in isaaclab/apps/isaaclab.python.headless.kit:

...
[settings] 
# MGPU is always on, you can turn it from the settings, and force this off to save even more resource if you 
# only want to use a single GPU on your MGPU system 
# False for Isaac Sim 
renderer.multiGpu.enabled = true 
renderer.multiGpu.autoEnable = true 
rtx.verifyDriverVersion.enabled = false # <- added it here 
'rtx-transient'.resourcemanager.enableTextureStreaming = true 
app.asyncRendering = false 
...

However, this didnā€™t work either, and I still get the ERROR_INCOMPATIBLE_DRIVER error.

Additional Context

Iā€™m training on a SLURM CentOS 8.1 cluster using Isaac Labā€™s cluster training setup. My local machine (Ubuntu 22.04) has a compatible driver version (535.247.01) so the error doesnā€™t come up when I train on that instead.

2

Thank you for your interest in Isaac Lab. To ensure efficient support and collaboration, please submit your topic to its GitHub repo following the instructions provided on Isaac Labā€™s Contributing Guidelines regarding discussions, submitting issues, feature requests, and contributing to the project.

We appreciate your understanding and look forward to assisting you.

1 Like