NUMA optimization?

During the webinar we were shown a quick few slides about NUMA utilization,
is there any additional information on optimizing this?
More specifically, how would one make sure both ram and gpu slots are assigned to the correct socket?

Additionally, any copy of the webinar sheets available?

There is also topic on vGPU optimization that has a link to a recording that covers NUMA utilization and shows how you would configure XenServer/XenDesktop with vGPU.

Erik Bohnhorst | Solution Architect – GRID
NVIDIA Corporation