Nvidia famously “unlaunched” the much-derided RTX 4080 12GB GPU a few weeks after it was announced. Instead, it chose to only move forward with its big brother, the RTX 4080 16GB. Despite both GPUs sharing the “4080” moniker, they have drastically different specs. Nvidia had said this was the reason for the confusing unlaunching. Now it appears the GPU will be reborn as the RTX 4070 Ti, according to regulatory filings from Gigabyte, and its AD104 die has also been photographed.
Gigabyte filed paperwork for its upcoming GPUs with the Eurasian Economic Commission (EEC), as reported by Videocardz. Although the card’s full specs aren’t listed, the titles include names such as GV-N407TEAGLE-12GD, so there’s no secret it’s a 4070 Ti. Overall Gigabyte will be offering a full suite of cards with eight models. This includes standard and “OC” variants of its Eagle, Gaming, and Aero, with the Aorus Elite and Aorus Master cards at the high end. Gigabyte usually offers a liquid-cooled model as well, but it’s not included in the EEC listing. The AD104 was built with TSMC’s 4N process. It’ll feature 7,680 CUDA Cores, 12GB of GDDR6X memory, 240 tensor cores, 60 ray tracing cores, and a 192-bit memory bus. This is one of its biggest differentiators from the 16GB card, which has a 256-bit memory bus. The RTX 3070 Ti also has a 256-bit bus.
The AD104 die is estimated to be 295mm², roughly half the size of the 608mm² AD102 die in the RTX 4090. The RTX 4080 16GB uses a 379mm² die named AD103. Prior to Ampere, the x104 dies were used in x80 GPUs, so this is a return to form for Nvidia. With Ampere, The RTX 3080 used the same AD102 die as the flagship RTX 3090, which is why there were so close in performance. For Ada Lovelace, Nvidia seems intent on putting more distance between its GPUs. Also, the AD104 die is about 100mm² smaller than its 392mm² Ampere equivalent.
It’s likely the reborn GPU will be launched at CES in January 2023. The timing is interesting as its main competitors from AMD will be launching on Dec. 13. This will give Nvidia plenty of time to figure out where its GPUs should land in pricing. AMD’s 7900 XTX and 7900 XT GPUs are $999 and $899 respectively. The company has also stated on the record that RTX 4080 is the target. Nvidia had previously priced the 4080 12GB at $899, so it’ll be interesting to see if it sticks with that price tag. The RTX 3070 Ti Founders Edition cost $599 at launch (theoretically).
Nvidia is in a bit of a sticky situation with its GPU pricing. Nobody is complaining about the $1,599 RTX 4090 as it’s easily the most powerful GPU ever made. Even AMD has admitted it has no answer for it. The stepped-down RTX 4080 16GB is seen mainly as a flop, with inventory languishing on store shelves weeks after it launched. This is despite reports of a smaller launch inventory than the 4090. For anything other than flagship GPUs, customers are much more sensitive to price now, and the ongoing economic situation hasn’t helped matters. Nvidia is obviously aware of this, so there’s hope it could price the 4070 Ti competitively. To us, that sounds like a $799 or even $699 GPU, but we’ll have to wait until January to find out.
Now Read:
- Class Action Suit Against Nvidia Filed Over Melted RTX 4090s
- Nvidia Is Reportedly Ending Production of Its Most Popular Turing GPUs
- Nvidia RTX 4090 Sets New Records in Password Cracking Benchmarks