Difference between DOCA DMA and RDMA

Hi, I have a question about DOCA DMA.
I’m trying to implement DMA between a DPU(BlueField3) and GPUs(A100) using DOCA DMA with nvidia-peermem. This approach shows the same resource usage and performance as when using RDMA with nvidia-peermem. Are there any system-level or hardware difference between these 2 methods?

Thank you!

hi

DOCA DMA provides an API to copy data between DOCA buffers using hardware acceleration, supporting both local and remote memory regions.

RDMA is a network procotol for exchange data between node.

Such 2 technology will have different implement under different scenario

Thank you
Meng, Shi