The latest news from @Kopite7kimi says that NVIDIA may equip the GeForce RTX 4080 with 16GB @ 23Gbps GDDR6X memory and the entire card (TBP) may consume up to 340W. In addition to the memory and power supply, most of the rest of the specifications of the explosion almost unchanged. But after upgrading from GDDR6 to the faster GDDR6X, the design power consumption has also gone up by 20W.
(via WCCFTech)
GeForce RTX 4080 Expected specifications.
" ● First, the card is expected to have a streamlined AD103-300 GPU configuration with 76 of 84 SM groups / 9728 CUDA cores enabled, compared to the previously expected 80 SM groups / 10240 CUDA cores. ● Second, the full GPU spec has 64MB L2 cache / 224 ROP grid cells, but the non-full-blooded version may only have 48MB L2 cache / lower ROP grid cells."
On the PCB side, the GeForce RTX 4080 is expected to use the PG136/139-SKU360 design.
"● Memory is 16GB @ 256-bit 23Gbps GDDR6X with a bandwidth of 736 GB/s -- slightly less than the 10GB @ 320-bit RTX 3080 (760 GB/s bandwidth). However, to make up for the memory bit width/bandwidth shortfall, NVIDIA is said to be using memory compression technology for the RTX 40 series Ada Lovelace product line."
Power consumption, @Kopite7kimi again updated to 340W, compared to the 320W mentioned earlier in the news increased by 20W, and 350W RTX 3080 in the same class.
As for the other RTX 40 graphics, SKU will also be more GDDR6X memory, it is not known. But it is certain that Micron has already started full mass production of 24Gbps GDDR6X.
Finally, referring to NVIDIA’s marketing strategy for previous generations of products, the RTX 4080 / 4070 will be the first model to launch to gamers outside of the RTX 4090.
If all goes well, RTX 4090 will arrive first around October 2022. As for RTX 4080 / 4070 whether to wait until 2023, is still to be decided by NVIDIA itself.
