There is no card in the Radeon RX 7000 line of graphics cards that can compete with the GeForce RTX 4090. However, it could be there. AMD says it could have developed a graphics card comparable to Nvidia’s flagship, but didn’t.
In an interview with ITMedia, AMD EVP Rick Bergman and AMD SVP David Wang discussed a variety of topics, including the one discussed above. Bergman stated that the RDNA 3 architecture allowed AMD to create an even more powerful 3D card than the RX 7900 XTX, but AMD followed a different strategy when creating a new line. Such an adapter would consume too much power and be too expensive.
It’s technically possible to design a GPU with specs that rival Nvidia’s flagship. However, a GPU designed in this way would be marketed as a 600W TDP graphics card with an estimated price of $1,600. After thinking about it, we decided not to adopt such a strategy.
The RDNA 3-based Radeon RX 7900 XTX is targeting a $1,000 price tag, which is considered the “top price” intended by high-end users among casual PC gamers.
The pricing strategy is the same as RDNA 2, with the top models Radeon RX 6900 XT and Radeon RX 6800 XT priced at $1000 and $700 respectively. However, the target price changes for each GPU generation.
We use this strategy to fit into the core infrastructure used by today’s gaming enthusiasts. At the same time, it should be possible to use existing power supplies, relying on common sense, be able to cool the inside of the case in terms of common sense. Radeon RX high performance products have been designed with these requirements in mind.
That is, AMD simply did not want to release a very expensive product, keeping prices at the level of the previous generation. However, the issue of energy efficiency was also important, and here RDNA 3, contrary to the promises of AMD itself, is approximately at the level of RDNA 2, and this is no longer enough to compete with Ada Lovelace.
The topic of chiplet GPUs was also touched upon. You can recall that initially absolutely all the rumors and leaks said that the Navi 31 GPU would be dual-chip. Perhaps AMD planned it that way, but eventually changed the concept. And now it became clear why.
As you know, modern high-performance GPUs contain more than 10,000 processing cores. This is more than 1000 times the number of processor cores. When trying to implement interconnects, the GPU does not work because the number of connection points will be huge, and reliable transmission of the electrical signal cannot be guaranteed. So, at the moment it is difficult not only in terms of man-hours, but also in terms of cost, to connect GPU dies in the same way as CPU dies. Nothing is impossible, but… Instead, it’s more efficient and less costly to build a large-scale GPU for now. Therefore, in the current generation, we decided to postpone the implementation of a multi-chip crystal
