RTX 5080 24GB and the future of the RTX 5090

BenchLife reports that NVIDIA would redirect future GB202 chip allocations towards RTX Pro cards. Partners manufacturing GeForce RTX 5090 cards would therefore no longer receive these chips to continue production. The outlet also mentions the Chinese RTX 5090 D and D V2 variants.

The distinction matters: BenchLife presents this as a reallocation of production, rather than a formal end-of-life announcement. Restricting GPU deliveries to manufacturers does not mean that all cards already produced immediately disappear from shops.

MEGAsizeGPU, who uses the @Zed__Wang account, and @hongxing2020 are two accounts sharing unofficial information about PC components. Their posts regularly fuel hardware rumours; they are not NVIDIA announcements.

On October 10, MEGAsizeGPU claimed that future GB202 supplies would be reserved for RTX Pro cards. Hongxing2020 instead used the term “EOL”, meaning end of life, and mentioned a 24GB RTX 5080. Their wording is therefore not quite equivalent.

The original posts are shown below so readers can check directly what each source claims.

According to BenchLife, this variant would replace the 16GB RTX 5080 in the first quarter of 2027. At the time of the report, however, partners had reportedly only received verbal notice that development documents would arrive soon, rather than the documents themselves.

This is a key point when assessing how far the project has progressed. A timeline discussed with partners is neither a final specification sheet nor a firm retail launch date. None of the official announcements consulted settles its name, price or a possible SUPER designation.

The current RTX 5080 officially has 16GB of GDDR7. Moving to 24GB would add 8GB, or 50% more capacity. That calculation concerns the amount of memory; it does not describe a 50% increase in graphics performance.

The potential benefit would mainly concern situations where available capacity becomes a constraint: large textures, complex creative scenes or certain local AI workloads. When the data already fits in the card’s memory, adding capacity does not automatically speed up the compute units.

The full configuration of this possible variant will therefore matter. The GPU, clock speeds, memory speed and power consumption are needed to compare two cards: the “24GB” label is no substitute for those details or for testing.

If RTX 5090 supply were to shrink for an extended period, the market would rely more heavily on remaining stock. But the sources consulted do not establish when it might run out or allow us to calculate a price increase in France.

An RTX 5080 with more memory could broaden what this model can do. Simply adding VRAM would not turn it into an RTX 5090, however. Readers weighing the two families should keep memory capacity, compute performance and availability distinct.