NVIDIA GPU Passthrough
I am working with a ros app on my stx deployment and i saw that it will need gpu access. It wasn’t clear for me on the documentation if the kubernetes pods have default access to gpu resources (its a AIO-Simplex), or if i need to configure the nvidia gpu operator. The problem is that from what i saw, the support was droppend since R8.0. I tried to install it pushing the chart to the starlingx repo but it fails to pull images from the nvidia endopoint. Any solutions on how to implement passthrough for kubernetes?
Hello, As you've noted, the NVIDIA GPU operator support was dropped a while back -- in late 2022. At this point, there is nobody in the community who is actively working on re-introducing this support. However, it's being considered for Q3-2026 for stx.13.0. Regards, Ghada stx releases PL -----Original Message----- From: joaopedrosnd@gmail.com <joaopedrosnd@gmail.com> Sent: Wednesday, December 17, 2025 11:36 AM To: starlingx-discuss@lists.starlingx.io Subject: NVIDIA GPU Passthrough CAUTION: This email comes from a non Wind River email account! Do not click links or open attachments unless you recognize the sender and know the content is safe. I am working with a ros app on my stx deployment and i saw that it will need gpu access. It wasn’t clear for me on the documentation if the kubernetes pods have default access to gpu resources (its a AIO-Simplex), or if i need to configure the nvidia gpu operator. The problem is that from what i saw, the support was droppend since R8.0. I tried to install it pushing the chart to the starlingx repo but it fails to pull images from the nvidia endopoint. Any solutions on how to implement passthrough for kubernetes?
participants (2)
-
joaopedrosnd@gmail.com
-
Khalil, Ghada