Usually, when a graphics card launches with two reminiscence configs, they’re basically the identical, simply with completely different reminiscence densities. That is the case with the GTX 580 3GB, the RX 580 8GB, and the uncommon GTX 960 4GB. That is not the case with the two fashions of the RTX 3080, and it seems just like the RTX 4080 may launch in an analogous state.
New leaks from MEGAsizeGPU (@Zed__Wang on Twitter) promise that the GeForce RTX 4080 will are available in two completely different variations: one with a 10-layer PCB sporting 12GB of GDDR6X reminiscence, and one with a 12-layer PCB outfitted with 16GB of the identical. In response to the leaker, the 2 variations have “fully completely different PCB designs,” which is uncommon for this type of factor.
Slightly than NVIDIA attempting to decide on which mannequin to launch, MEGAsizeGPU says that the 2 variants will in actual fact be launched and launched collectively, “similar to 1060 6G and 1060 3G”. He’s in fact referring to the GeForce GTX 1060, which launched with 3GB and 6GB fashions. Opposite to what you would possibly anticipate given the easy doubling of reminiscence capability, these Pascal GPUs really differed barely of their capabilities, with the GTX 1060 6GB card having extra shaders and texture items in comparison with the 3GB mannequin.
It is doubtless that the identical shall be true of those GeForce RTX 4080 playing cards, as properly. 4GB of reminiscence is rather a lot to trim from a GPU, and it is doubtless that the 12GB card will see a discount to a 192-bit reminiscence bus width from the complete 256-bit width of the AD103 GPU. Ordinarily we would anticipate that to end result within the card being slapped with the “x70” title, however supposedly each of those fashions shall be branded “RTX 4080.” Fairly complicated.
Much more complicated is that common leaker kopite7kimi says that the RTX 4070 and RTX 4080 12G fashions may share a PCB. If that is the case, then both the AD103 and AD104 GPUs are pin-compatible, or the RTX 4080 12G card shall be utilizing the smaller AD104 GPU. That truly is smartāAD104 has a 192-bit reminiscence bus, and as a smaller GPU, yields on it should doubtless be higher. It is unattainable for anybody however NVIDIA to say at this level, although.
Earlier, Japanese Twitter person Elysian Realm (@KittyYYuko) shared an image of what they purport to be a GeForce RTX 4080 Founders’ Version card. The image will not be notably high-resolution and the cardboard is shrouded in an anti-static bag, however you may clearly see a cooler that strongly resembles the GeForce RTX 3090’s inventory cooler with “RTX 4080” emblazoned on it.
Elysian Realm tweeted again in Might that the RTX 4080 FE “seems barely larger than the RTX 3090 FE.” This newest picture would appear to corroborate that assertion, because the cooler seems to be extraordinarily tall to accommodate what seems like a 120-mm (or bigger) fan. Responding to a query, the leaker confirms that the cardboard occupies 3 enlargement slots, similar to the RTX 3090.
