@SamJWasserman@nvidia@NVIDIAAI@SamJWasserman , even on the nvidia page documenting clustering they have mentioned 400G cable. I know NICs are capped at 200G, will the above cable work ?
I thought QSFP 112 is the only varient should be bought
@SamJWasserman@nvidia@NVIDIAAI 0.5 m not enough for closing the loop ?
And what I read so far is that tp=3 < tp=2 because of this topology.
Did you find any use case with 3rd dgx ?