Hi, just wanted to know if there is any plan to get vgpu support for Jetson ?
While I don’t think official Nvidia vGPU support currently exists for the Tegra chips, the possibilities are endless. A more powerful machine like the Jetson AGX or the future Orin AGX would much benefit from the ability to run multiple hardware-accelerated VMs simultaneously for various enterprise and educational applications, allowing companies and schools to split up and provide SSH access to these devices. Students and employees could test their Tegra code or other generic CUDA-accelerated applications on a Jetson cluster, with more people than devices.
We would like to get more detail about the use-case. For vgpu, do you mean connect a desktop GPU card to PCIe slot on Xavier developer kit and use it for running deep learning? This is not the case supported in current Jetson releases.
Nope–the Jetson boards have pretty great integrated GPUs for their price and size, and the use case we imagined was more of splitting that GPU up into mdev’s. Then, multiple VMs for different users or for testing could be used on the same machine, maybe even simultaneously. The vGPU software Nvidia provides allows this functionality on desktop cards but we were looking for this same functionality on the Tegra-based dev boards.
Makinbacon and I are part of the same development team, so as he said, we think this feature would be great on the Jetson products given the increasing power and the growth of arm64 devices in different projects.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.