Clarification on the purpose of multiple ConnectX-8 ports on MGX ARC Pro

Accelerated Computing Aerial Forum
,
1

Hi NVIDIA team,

I would like to better understand the intended architecture behind the large number of ConnectX-8 Ethernet ports on the MGX ARC Pro platform.

Is the main purpose of these multiple ports to allow ARC Pro to directly connect to multiple O-RUs, potentially eliminating the need for a dedicated fronthaul switch in small deployments?

For example, could a topology like this be officially supported?

PTP/SyncE Grandmaster ──> ConnectX-8 Port A
O-RU                  ──> ConnectX-8 Port B

with the O-RAN 7.2x C/U-plane running directly between ARC Pro and the O-RU.

Or are the multiple ports mainly intended for:

  • Multiple O-RUs / multiple cells
  • Separate fronthaul and backhaul interfaces
  • Redundant fronthaul paths
  • RU emulator / test equipment connections
  • Future multi-node or AI-RAN expansion

while a dedicated fronthaul switch is still expected to provide the PTP/SyncE timing and L2 switching fabric?

In particular, can ConnectX-8 on ARC Pro perform timing-related functions between ports, such as PTP Boundary Clock / Transparent Clock or SyncE clock distribution, or should each O-RU receive timing independently from a GM or fronthaul switch?

It would be very helpful if you could clarify the intended role of the multi-port ConnectX-8 design and provide a recommended switch-less topology, if supported.

Thanks.

2

Since ARP Pro is designed for the D-RAN deployment, ARC Pro systems from some OEM vendors, such as QCT, will equip a GNSS receiver port and 1 PPS in/out ports. Also, CX-8 has PHC and 1 PPS in/out ports. So, it supports the LLS-C1 synchronization topology.

Thank you.