It looks like Nvidia has changed the RTX 40 series line-up once again. The board SKU we thought to be the RTX 4070/Ti has reportedly been renamed to RTX 4080 12GB. Moreover, the maximum TGP and clock speeds of the RTX 4090 and 4080 GPUs have leaked.
The leaker kopite7kimi and VideoCardz shared a summary showing the three upcoming GPUs and their respective specifications. As previously reported, the RTX 4090 should feature the AD102-300 GPU with 16,384 CUDA cores that can boost up to 2,520MHz. The memory subsystem is rumoured to feature 24GB of GDDR6X memory at 21Gbps. Default TGP should be around 450W, but this value could be configured via BIOS to go as high as 660W.
For the RTX 4080 16GB model, we have the AD103-300 GPU with 9,728 CUDA cores boosting up to 2,505MHz. As the name suggests, it packs 16GB of GDDR6X memory clocked at a whopping 23Gbps (you read that right). Based on the specs of the other GPUs, this is the only one running memory at this speed. The TGP is rumoured to go up to 516W, but the default value is “only” 340W.
Lastly, the RTX 4080 12GB model will pack the AD104-400 GPU with 7,680 CUDA cores and a boost clock of 2,610MHz. With 12GB of GDDR6X memory running at 21Gbps, this GPU features a TGP of 285W that can go up to 366W.
Discuss on our Facebook page, HERE.
KitGuru says: Considering we're just six days away from the reveal of the RTX 40 series, the GPU line-up and specifications should be pretty similar to what we have here. As for the RTX 4070 GPU, it looks like Nvidia chose the 10GB SKU with 7,168 CUDA cores for the upper mid-range card.