The reason for this is said to be the insufficient size of the graphics chip that AMD decided to use in its solutions.
Insider @Kepler_L2 noted that AMD will continue to use a monolithic GPU design instead of the more efficient chiplet design used in current RDNA 3 series cards, such as Navi 31 and Navi 32.
This approach limits the ability to increase the die area, which affects overall performance.
The UDNA architecture will be based on TSMC’s 3nm process (N3E) and will be used in both consumer graphics processors and Instinct MI400 series compute cards.
However, experts say that AMD will need a significant performance increase to compete with NVIDIA in the premium graphics market.
