GPU comparisons for local AI

Two cards, side by side: which fits bigger models, which runs them faster, and the speed of every popular open model on each.