A Twitter user named @Hongxing2020 recently shared photos of a mysterious MSI GeForce RTX 3080 Ventus 3X 20G OC, packing an impressive but unusual memory capacity of 20GB. The user claims that up to 100 units are available for between $432 to $576, insinuating that these could be leftover cards from mining operations. Core specifications are unknown, but the graphics card is likely the long-rumored 20GB variant of the vanilla RTX 3080, modified to support Micron memory chips with double the capacity.

It wouldn't be the first time Nvidia has reportedly messed around doubling VRAM capacities on its high-end GPUs. Almost every generation of Nvidia graphics cards has featured some rumor or suspicion pointing to a 2x higher capacity variant well after the GPU's initial release.

In the past, Nvidia has sold different memory capacity variants of the same GPU model, with one variant featuring 2x the memory capacity over the other. This strategy was prevalent during the Kepler generation and older models, where cards like the GTX 780 3GB and GTX 770 2GB had additional 6GB and 4GB models, respectively.

Making a new GPU with 2x the memory capacity isn't hard; it's one of the easiest ways to add additional VRAM capacity to a GPU. All Nvidia has to do is swap out memory ICs with ones featuring two times the capacity of the previous variant. Adding other levels of VRAM lower or higher than 2x on the other hand gets much more complicated. That would involve modifying the memory bus width and memory configuration, as memory manufacturers almost always make memory ICs based on 512MB, 1GB, or 2GB memory configs with no in between capacities.

More often than not, higher memory configurations come on higher tier products, like the RTX 3090 and RTX 3090 Ti with 24GB of VRAM. However, the RTX 3060 comes with 12GB of memory, even though the faster RTX 3060 Ti, RTX 3070, and RTX 3070 Ti have 8GB. That's perhaps because Nvidia deemed 6GB too little for a modern RTX 3060, unlike the previous generation RTX 2060. But the 3060 has a narrower 192-bit bus, where the 3060 Ti, 3070, and 3070 Ti have a 256-bit bus.

If Nvidia was thinking about an RTX 3080 20GB, it likely didn't make sense. The RTX 3080 Ti has 12GB, and later we got the RTX 3080 12GB. Nvidia seems to have gone with more memory bandwidth rather than more memory capacity for the current RTX 3080 series.

The RTX 3080 12GB wasn't a straight upgrade in memory, either. Nvidia had to upgrade the RTX 3080 12GB with additional memory controllers, and it enabled a few more more CUDA cores (two SMs) to go along with the extra 2GB. While the core configuration is on paper only 3% faster than the RTX 3080 10GB, the additional 20% memory bandwidth proved very beneficial, potentially matching the RTX 3080 Ti (depending on core clocks).

An RTX 3080 20GB likely would have been a downgrade for gamers, as it would have remained with a 320-bit bus. Another issue with a 20GB model is memory utilization in games. Only a select few titles running at very high resolutions and maxed-out texture packs can break 12GB of VRAM. As a result, the RTX 3080 20GB's additional 8GB of VRAM would rarely be helpful for most gamers.

A 20GB SKU might be a good upgrade over a 10GB model, but with a 12GB variant already existing, a 20GB model for gamers makes no sense. The only place an RTX 3080 20GB would make sense is in the prosumer side of the market, where gobs of memory capacity are beneficial for non-gaming tasks such as 3D rendering and high-resolution video timelines with tons of different visual effects. But Nvidia already has the RTX 3090 and RTX 3090 Ti to bridge this gap, making the RTX 3080 20GB even more pointless.