For starters the 2 dies which have appeared have codenames like GA103 and GA104, standing for RTX 3080 and RTX 3070 respectively. Perhaps the largest shock is the Streaming Multiprocessor (SM) rely. The smaller GA104 die has as a lot as 48 SMs, leading to 3072 CUDA cores, whereas the larger, oddly named, GA103 die has as a lot as 60 SMs that lead to 3840 CUDA cores in whole. These enhancements in SM rely ought to lead to a notable efficiency enhance throughout the board. Alongside the rise in SM rely, there’s additionally a brand new reminiscence bus width. The smaller GA104 die that ought to find yourself in RTX 3070 makes use of a 256-bit reminiscence bus permitting for 8/16 GB of GDDR6 reminiscence, whereas its greater brother, the GA103, has a 320-bit vast bus that enables the cardboard to be configured with both 10 or 20 GB of GDDR6 reminiscence. In the pictures beneath you possibly can try the alleged diagrams for your self and see if this appears pretend or not, nonetheless, it is suggested to take this rumor with a grain of salt.
