The RTX 4080 might be subject to as many memory switcheroos as its forebearer

pcgamer.com:

Has there been a GPU generation with such fiercely debated and rumoured specifications as Nvidia's RTX 40-series(opens in new tab)? I'd say no, there hasn't been. The upcoming lineup of graphics cards has shapeshifted far too many times to count over the past year. However, with a potential launch window coming up fast, we're beginning to see these cards' final form—and there could be a few different memory configurations on the way.

We're expecting the RTX 4090 to lead the charge for the RTX 40-series, which is likely to feature 24GB of GDDR6X memory as standard. That's not particularly groundbreaking information as the current generation RTX 3090/Ti also come with as much.

However, it's the RTX 4080(opens in new tab) that could be a more interesting sight at its release, though when that might come is not entirely clear. According to leaker MEGAsizeGPU(opens in new tab) (via Videocardz(opens in new tab)), where we'd expect a single graphics card, we may find two. The RTX 4080 is rumoured to arrive with two different memory configurations: either 12GB GDDR6X or 16GB GDDR6X.

Even more surprising is that these are said to be fundamentally different graphics cards. The 16GB variant is reported to be using a 12-layer PCB and the 12GB variant a 10-layer PCB. And both are said to be arriving at the same time.

Nvidia has offered such a split in the past, though arguably where it made more sense in the more entry-level market with the GTX 1060 6GB/3GB. At the higher end of the performance spectrum, you'd suspect Nvidia to be a little more sure of what's required of its high-end cards.

Though competition will do things to a graphics card company, and it wouldn't be entirely unlike Nvidia to float the idea of multiple memory