Difference between ConnectX-6 DX and ConnectX-6 VPI cards

Hello everyone,

I’m looking into buying a ConnectX-6 NIC and i’m wondering about the different product lines and SKUs.

What specifically are the differences between the different products?
I’m for example comparing MCX653106A-ECAT to MCX623106AN-CDAT.

The first one is a IB/Ethernet card from the ConnectX-6 series and the other one a ethernet card from the ConnectX-6 DX series. Both have 2 QSFP56 Ports capable of speeds up to 100 Gbit/s. Are the only differences the supported protocols and can i expect both cards to behave the same in ethernet mode?

I’m also interested in using their RDMA features with DPDK or Doca GPUNetIO. The GPUNetIO documentation states the following: “It is highly recommended to only use NVIDIA adapter from ConnectX-6 Dx and later.”. Does this include the ConnectX-6 VPI models in Ethernet mode?

Thanks in Advance,
Michael

Hello michael.f,

Thank you for posting your inquiry on the NVIDIA Developer Forum - Infrastructure and Networking - Section.

No, ConnectX-6 VPI models do not have the same offload capabilities as the ConnectX-6 Dx. VPI models have the capability to change the port-type to Ethernet or Infiniband.

You can review the differences between them through the following link → Networking Adapters - NVIDIA Docs

Thank you and regards,
~NVIDIA Networking Technical Support

Thanks for the reply,

Could you please elaborate on the different offloading capabilities?

Both manuals list PeerDirect and RDMA over Converged Ethernet as supported features described in the same wording.
Connectx 6 Dx Manual, Connectx 6 VPI Manual

Thanks,
Michael