haxxiy on 18 February 2023
JEMC said:
If I remember correctly, you follow this news more than I do, the node Nvidia used is nothing more than a more refined 5nm one, not a "true" (if that term can even be used) new node. I can't see this being the only reason for the improved efficiency Ada shows when dialed back to its most efficient point. |
True, it's the same node density-wise, but Nvidia must be paying an extra for something, which might be fine-tuning for their architecture yielding a better frequency-voltage curve. Definitely not one that explains entirely the better result than AMD's, but maybe halfway there.
Essay Pro