Cuda streams on Tesla v100

Assuming that I’m not using very much memory, how many CUDA streams can run concurrently on one Tesla V100? Is there a hardware limit aside from memory?

Thanks, Roger