Problems using NVIDIA GPUs on AWS & Azure while on local works fine

We have a kubernetees implementation with GPUs that is working fine on local machines (quadro 4000 GPU) but when running on AWS or Azure with Tesla V100, the performance is 1/4 of what is on the local machine. the cloud machines are by fast stronger than the workstations, and still performance is bad. Not sure this is the right place to post this question, but what would be the right way to get this fixed ?