Recent leaks indicate that NVIDIA may be preparing a second RTX 5080 variant than what is known thus far (from leaks), per Wccftech. The Blackwell-based GPU will reportedly feature a second variant with 24GB of VRAM, which can be done by packing 3GB GDDR7 modules.
RTX 5080 24GB

This information can be traced back to Chiphell forums, where user PolyMorph simply stated:
(RTX) 5080 will later come with 24GB (variant)
According to Wccftech, the reason why RTX 5080 may first release with a 16GB variant as leaks earlier suggested may be due to the limited yields of GDDR7 modules. A typical high-end GPU features eight memory modules (some ultra-enthusiast models have more), and the density of these chips determines the total VRAM capacity of the entire card. Using 2GB modules means you’ll get 16GB in an eight-chip configuration, and by using 3GB modules instead, you’ll get 24GB in total.
As a quick refresher, the RTX 5080 is rumored to feature 10,752 CUDA cores (roughly half of the RTX 5090), paired with 16GB GDDR7 RAM in a 256-bit memory bus, and consumes up to 400W of power. This card is also expected to surpass the RTX 4090 in raw performance.

One thing to note is the naming: while it’s still very early to discuss product names, NVIDIA has already set many precedents on adding different variants on the same GPU model which can have varying performance. Team Green tried this with the RTX 4080 16GB and 12GB, where the latter is significantly worse in performance despite sharing the same name – this was later reversed and the GPU has since became what we know as RTX 4070 Ti today.
It’s unclear if the 24GB variant may be called RTX 5080 with no differentiators in naming, or it’ll simply be part of a mid-cycle refresh (as the leaker seem to indicate) that will be part of the RTX 50 SUPER series. In any case, take these information with a good amount of salt.
Pokdepinion: I certainly hope the naming doesn’t get worse.