RTX 5080 24 GB: replacement plans take shape, January remains uncertain

The latest development comes from BenchLife. In its article of 2 November, the Taiwanese outlet reports that NVIDIA's partners had still not received the RTX 5080 24 GB design documents by the end of October. These “design kits”, expected that same month, are needed to develop their cards.

Its sources therefore consider a January release difficult to achieve. Tight supplies of printed circuit boards and components would also complicate launch preparations. BenchLife does not, however, provide a new date.

An unveiling at CES followed by later availability remains a possibility raised by Wccftech, not a schedule announced by NVIDIA. These two stages must be distinguished: unveiling a card does not mean it is immediately on sale. At this point, talk of an official delay would be premature, since the manufacturer has announced neither this variant nor its release date.

The initial scenario was reported on 30 October by VideoCardz, drawing on information from ChannelGate, associated with Board Channels. It describes a 24 GB RTX 5080 expected between December 2026 and January 2027, which would replace the 16 GB model. The withdrawal of the RTX 5090 D V2 is also mentioned in this reshaping of the Chinese product range.

That geographical scope matters: the report does not confirm the same transition in France or elsewhere in the world. ComputerBase also highlights this limitation. Likewise, seeing the information on several sites does not amount to as many independent confirmations: several publications refer back to the same Chinese source.

ChannelGate also claims that NVIDIA has exhausted its GPU stocks for the RTX 5080 16 GB, while cards already in the distribution channel would allow roughly another two months of sales. This estimate is neither a verified global inventory nor proof of an immediate shortage in French shops.

The account @harukaze5719 shared this launch window on 30 October:

These latest developments build on the first reports on the RTX 5080 24 GB that we covered in October, with an additional caveat over timing.

The proposed capacity increase is easy to quantify: an extra 8 GB, or 50% more memory. Eight 3 GB chips would provide 24 GB, versus 16 GB with eight 2 GB chips. Reports discuss this approach alongside retaining a 256-bit bus. NVIDIA has yet to confirm this configuration.

The necessary technology already exists. Samsung unveiled 24-gigabit GDDR7 chips in October 2024, equivalent to 3 gigabytes per chip. The manufacturer claimed 50% higher density than its 16-gigabit chips in the same package size. Pay attention to the units: a 24 Gb chip holds 3 GB, not 24 GB.

Enlarge

This announcement demonstrates the feasibility of higher-density memory. It does not prove that Samsung will supply this possible RTX 5080, or that its advertised maximum speeds will be used on the card. The supplier, memory data rate and final configuration remain to be established.

The official specification sheet for the current RTX 5080 provides a solid basis for comparison. The entries in the 24 GB column remain unofficial information or unknowns; they do not form a final product specification.

NVIDIA also lists a 2.30 GHz base clock and a recommended 850 W system power supply for the current reference RTX 5080. These figures must not automatically be attributed to the future variant. Custom models may also have different clock speeds.

The potential benefit concerns workloads that run up against memory capacity limits. In games, more VRAM could provide extra headroom for large textures or certain demanding settings. In content creation and local AI, it could allow more data to remain on the GPU. These are possible consequences of increased capacity, not measured results from this card.

However, a game or application that already fits within 16 GB does not automatically become faster with 24 GB. Memory capacity, data rate and compute power are distinct characteristics. At the same data rate, increasing capacity alone is not enough to increase bandwidth.

Essential details are therefore still missing to assess this variant: the GPU used, clock speeds, power consumption, price and independent tests. A possible “SUPER” name is not confirmed either. Specifications from older RTX 5080 SUPER rumours cannot simply be carried over to these new reports.

The China-specific part of the report does not establish the end of the entire RTX 5090 family. On 16 October, Jeff Fisher told 4Gamers that the RTX 5090 was not discontinued and that NVIDIA would continue to ship it, while acknowledging supply chain constraints.

That statement does not guarantee plentiful stock, but it remains separate from rumours about chip allocations or the D V2 version. We previously covered NVIDIA's denial and the pressure on RTX 5090 stocks. Adding memory to an RTX 5080 would not, by itself, make it equivalent to an RTX 5090 anyway.