Nvidia is working on a new top-tier flagship GPU, known as the RTX 3090 Ti, and it will have even faster memory than its predecessors. Micron recently unveiled a super fast GRR6X memory built with Nvidia that can hit speeds of 21Gbps. That's 1.5Gbps more than existing 3090, and five Gbps more than AMD's best GPUs. The card is also expected to have a greater number of CUDA cores, more RT and Tensor cores -- though what clock speed, is anyone's guess.
Perhaps more dramatically, however, the card could draw as much as 100W more power than its predecessor, which doesn't seem like a fair trade considering the performance increase from the 3090 is unlikely to be super dramatic.
Nvidia's current lineup of RTX 3000 GPUs is topped out by the RTX 3090 and 3080 Ti, the latter of which is almost as fast as the 3090, except in a few niche cases. It's also usually a lot cheaper, making it the far better card for gaming, even if the 3090 does enjoy double the memory.
The 3090 Ti however, will add a little extra separation between the top gaming and workstation GPUs, with the 3090 sporting around 250 more CUDA cores than the 3090 -- and 500 more than the 3080. It'll also have that faster memory, which will give it around 70 GBps extra bandwidth that the 3090.
For RTX 3090 Ti pic.twitter.com/hkTlCrCIQ7
— Uniko's Hardware (@unikoshardware) November 29, 2021
In reality, all of this will probably equate to a few extra percent gaming performance, but it won'tbe much. In a trade off for an extra 100W of power (making this a 450W TDP card) it's probably not going to be worth it.
What is intriguing, however, is that Nvidia feels the need to do this at all. Does that suggest that AMD may have a refresh of its own top cards which could nip at Nvidia's heels that bit more? We'll have to wait and see.