Judging from these early rumors from Videocardz.com, it appears that Nvidia will not be joining AMD in the memory amount game, sticking to 12GB for its RTX 3080 Ti graphics card.
Based on the same GA102 GPU like the RTX 3080, the RTX 3080 Ti could end up with 80 GPU clusters enabled, featuring 10240 CUDA cores, and coming with 12GB of 19Gbps GDDR6X memory on a 384-bit memory interface, leaving it with 864 GB/s of memory bandwidth.
The RTX 3070 Ti will still use the GA104 GPU, maxing it out with a total of 6144 CUDA cores, but still come with 8GB of memory. We are likely looking at a faster 16Gbps GDDR6X memory on the same 256-bit memory interface.
As noted, the RTX 3080 Ti could launch in mid-April, while the RTX 3070 Ti could come by the end of May.