
NVIDIA's upcoming GeForce RTX 5090, reportedly launching in 2025, could feature a 512-bit bus and a massive 32GB of GDDR7 memory. This design choice would simplify GPU development and scaling across their lineup.
Today, a relatively reliable insider, @kopite7kimi, shared some highly intriguing information. According to their sources, NVIDIA's next top-tier graphics card could feature a 512-bit data bus between its video memory and GPU. The Californian semiconductor giant last used such an interface in its gaming solutions way back in 2008 (GTX 280-285).
By loading, you allow content from X (Twitter) (cookies may be set).
The so-called "Ada-Next," or GeForce RTX 5090, is currently slated for release no earlier than 2025 and is highly likely to feature GDDR7 video memory. Furthermore, if the 512-bit interface rumors prove accurate, the graphics card would boast a 32-gigabyte video buffer, a capacity that aligns perfectly with that bus width.
This seems like a pragmatic approach from NVIDIA. Had they chosen a 384-bit interface, boosting video memory beyond 24GB would have been a real headache for engineers. A 48GB jump is essentially overkill for current needs, and from a technological standpoint, remaining at 24GB for over four years just doesn't make sense.
Moreover, this opens up significant room for the green giant's engineers to experiment with lower-tier graphics cards. For example, "cutting" the bus from 512 to 384 bits could enable the release of a hypothetical RTX 5080 Ti, and reducing it to 352 bits could result in an RTX 5080.
Whether this information is confirmed or debunked, we'll only find out towards the end of this year or early next, as benchmark leaks from hypothetical GeForce RTX 5090 test samples begin to surface online.