We are in the process of planning the specifications and compatibility of how the power distribution will work with the devkit.
Note that for deployment, we anticipate embedded Quadro or Tesla MXM’s being deployed by users, so a customized or 3rd-party ecosystem carrier for Xavier module + MXM would be able to power the MXM site as it normally would per the MXM specifications.
If one were to build a board with the embedded Tesla MXM and provide power correctly, and connect it to the Xavier’s PCIe, would the CUDA drivers work? i.e. is it just a hardware issue or will the driver need to be updated as well?