The 32 CPU quote [originally from POF's blog and repeated w/o correction elsewhere] is incorrect; the HP server in question has 8 quad-core _CPUs_, for a total of 32 _cores_.
So still a factor of 2.5 away from our hypothetical 80 core requirement, and not a refutation of GP's claim that x86 maxes out at 8 sockets.
So still a factor of 2.5 away from our hypothetical 80 core requirement, and not a refutation of GP's claim that x86 maxes out at 8 sockets.