Those numbers suggest that they have caught Nvidia in performance per transistor. As for the die area being larger, I suspect that the larger memory bus might b

Those numbers suggest that they have caught Nvidia in performance per transistor. As for the die area being larger, I suspect that the larger memory bus might be partly responsible. The transistors used for IO stopped shrinking on new nodes a while ago, so they use plenty of die area.