NVIDIA is allegedly working on upgraded memory variants of the GeForce RTX 5070 and RTX 5080 graphics cards. The rumor comes from the Chiphell forums (via panzerlied), claiming that the Blackwell family might get larger memory buffers in the future. The SKUs mentioned include the RTX 5080 and the 5090.
- The GeForce RTX 5090 sits comfortably at the top with 32 GB of graphics memory, sufficient for the most demanding games in the near future.
- The GeForce RTX 5080, however, falls short in certain games like “Indiana Jones and the Great Circle” and “Ark: Survival Ascended” at 4K.
- The GeForce RTX 5070 is also VRAM bottlenecked in several games at 1440p
The NVIDIA GeForce RTX 5080 leverages 16 Gb GDDR7 chips clocked at 30 Gbps. Replacing them with 24 Gb dies will result in a 24 GB buffer. Most units come with a large overclocking envelope, and it won’t be a surprise if the clocks are also revved up on the 24 Gb memory dies.
Seeing as the RTX 5080 uses the fully enabled GB203 die, the shader configuration will probably remain unchanged. A switch to the GB202 die is possible but unlikely given the lack of proper competition from AMD.
The GeForce RTX 5070 Super, on the other hand, may increase the shader count to 6,400 to exploit the full GB205 graphics core. The memory buffer can be expanded to 18 GB using the newer 24 Gb modules. Memory clocks of 30 Gbps should be easily achievable.
Overall, you can expect mild performance and memory increases from the RTX 5070 Super, and a larger memory buffer plus lower pricing for the RTX 5080 Super.
