P2P DMA (or GPU Direct RDMA) between an FPGA and GPU

[samy012]

Hello,

I am inquiring about the feasibility of utilizing only PEG (PCI Express Graphics) of COM-e (Computer-on-Module Express) for Peer-to-Peer (P2P) Direct Memory Access ( or GPU Direct RDMA)  between an FPGA and GPU.

I am considering using Raptor Lake-P (H series) or Meteoh Lake(H series) . PEG specification and Connection Topology are as follows.

< PEG specification >
- Raptor Lake-P (H series) offers "PCIe Gen4 x8," "PCIe Gen4 x4," and "PCIe Gen4 x4".
- Meteoh Lake (H series) offers "PCIe Gen4 x8," "PCIe Gen4 x4," and "PCIe Gen4 x4".

< Connection Topology >
"PEG: PCIe Gen4 x8": between the COM-e Module and GPU.
"PEG: PCIe Gen4 x4": between the COM-e Module and FPGA1.
"PEG: PCIe Gen4 x4": between the COM-e Module and FPGA2.

I would like to know if this topology is capable of achieving P2P DMA (or GPU Direct RDMA). If you have experimented with a similar topology, could you please provide your insights? Additionally, is a PCIe Switch IC absolutely necessary for this setup?

Best regards,

[RamyerM_Intel]

Hello samy012, 

I have read your concern and I want to let you know that we have a dedicated forum for this type of issue, where you can get faster answers. I am moving this to the FPGA Forum for you.

Ramyer M. 

Intel Customer Support Technician