Two xaviers communicate through pcie, which affects the bandwidth of RGMII network interface

We have two xavier modules on the self-developed board. The two modules communicate with each other through pcie, and the pcie becomes a virtual network card. We use the iperf to test the network bandwidth. When the pcie network is full of bandwidth, the bandwidth of the RGMII native network card decreases from 950M to 100M. If the pcie network communication ends, the bandwidth of the RGMII native network card recovers to 950M. When the pcie network is tested, the CPU occupancy rate is very high, Is this normal? How to reduce the impact of pcie network on RGMII network


The above two pictures show the CPU utilization during the pcie network test

Is RGMII also a xavier-xavier connection?

No, the rgmii network is connected to an external router

Why didn’t anyone reply to me

This feels like a known limitation and we don’t have any solution for now. Will check internally.

Hi, What about this problem? Is there any solution?

Hi lixing.gao,

We can’t reproduce your issue.
Please share the detail steps. Thanks!

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.