Can I have A16 & L40 cards in the same node, is that a supported configuration? I realize that I cannot time sliced profiles that overlap virtual GPUs. But I’m just talking about using different physical cards in the same node. I see documentation pictures where nvidia-smi is showing different physical cards on the same node / host.
No, you cannot. Ampere and Ada uses different RM (software/hardware) and therefore this is technically not possible as the vGPU manager can manage the one or the other.