There’s going to be a beefier GeForce RTX 4080 on its means within the first few months of 2024, so is the most recent graphics card hearsay on the social grapevine. The identify, Tremendous or Ti, has but to be determined nevertheless it’s speculated to be utilizing a cut-down model of the AD102 GPU, that is the GPU at present powering the RTX 4090. Energy restrict is meant to be lower than 450W and the worth vary is just like the RTX 4080’s.
That is all in line with MEGAsizeGPU on Twitter (through Videocardz) however as with all publish on social media with exactly zero exhausting proof to again it up, it is best to deal with it as being not more than a bathe thought, no matter who really posts it. That is mentioned, there’s some advantage to breaking down what’s really being claimed.
Let’s begin with using the AD102. The present GeForce RTX 4080 makes use of the smaller cousin of Nvidia’s top-tier GPU, the AD103. That has a most of 80 SMs (Streaming Multiprocessors, which comprise all the shaders, tensor cores, and ray tracing items) and a reminiscence bus that is 256 bits in complete width.
Nevertheless, the model that is used within the RTX 4080 solely has 4 SMs disabled and the remainder of the die runs as regular. Which means that there’s little tangible profit to making a quicker product utilizing that chip, as a totally enabled AD103 operating on the identical clock pace because the one within the RTX 4080 would simply be just a few % quicker.
The AD102 has plenty of potential, although. That chip has as much as 144 SMs and the complete bus width is 384 bits. The GeForce RTX 4090 makes use of one with 128 SMs enabled, so Nvidia might use one with, say, 100 SMs for an RTX 4080 Ti. That is the chip that is already used within the RTX 5000 Ada Era workstation card and it runs with an identical clock pace to that within the RTX 4080, albeit with a a lot decrease energy cap.
Such a configuration might see the Ti model being as a lot as 25% quicker than the unique mannequin, particularly if the cardboard used a 320-bit reminiscence bus configuration and drops in 20GB of VRAM, too. The RTX 4080 hasn’t been a runaway success for Nvidia, as its worth is vastly greater than any earlier 80-class mannequin. The GeForce RTX 3080 launched with an MSRP of $799, whereas the RTX 4080’s was $400 greater. An up to date model of the Ada Lovelace card would not be cheaper but when the worth was the identical, or inside a handful of greenbacks, then a Ti variant would possibly fare higher.
Even so, I wrestle to imagine the market shall be all that . Whereas the RTX 4090 is ridiculously costly, there is a clear mindset amongst such customers that in the event you can afford an RTX 4080, you would possibly as effectively spend extra to get the 4090. It isn’t going to be any completely different with a 4080 Ti, particularly since Nvidia is not going to place its efficiency too near the 4090.
And this is only one a part of the entire RTX 40-series drawback: Nvidia put itself right into a place the place it will wrestle to do mannequin refreshes that may appeal to patrons, with enticing specs and worth factors, with out killing every thing that has gone earlier than. The Ti variants of the RTX 4070 and 4060 exist already, so what identify would Nvidia use for quicker variations of these?
The Tremendous moniker has solely been used as soon as earlier than, within the Turing collection, and was used as an alternative of Ti, other than the RTX 2080 which had each. If Nvidia does revive the identify, it’d trigger confusion amongst those that at present personal a Turing-based Tremendous mannequin and need to improve. If it would not and simply releases the refreshed model with the identical names because the previous ones, how would you simply inform them aside?
Nvidia might want to do one thing, just because its stock of Ada Lovelace GPUs shall be filling up with dies that at present haven’t got a selected mannequin to be housed in. All massive scale chips undergo a course of referred to as chip binning as soon as they have been made. This includes testing all of the sections for proper operation, in addition to operating them at quite a lot of clock speeds and voltages.
Those who fail to fulfill the necessities for a present mannequin, however are nonetheless completely usable, are positioned in storage till they are often utilised elsewhere. Up to now, Nvidia did this by utilizing them in Ti or Tremendous variations or by placing high-end chips into lower-end fashions and disabling numerous elements, with out altering the product identify.
However we have already acquired RTX 4060/4070 Ti fashions in the marketplace, so if Nvidia does launch an RTX 4080 Ti within the first quarter of 2024, what else is left to be launched? It might spoil us with an RTX 4050 (please do not) however that is the one phase that does not at present have a 40-series mannequin. Perhaps it’s going to give us an RTX 4070 Ti Tremendous that sports activities 16GB of reminiscence utilizing the AD103 GPU and a 256-bit bus. Could be enjoyable to see if Nvidia can match all the moniker onto a Founders Version shroud.
For subsequent yr, it will be refreshed variations or nothing in any respect, with Nvidia’s subsequent era of GPUs not anticipated to reach till effectively into 2025. Clearly, that is all right down to Nvidia to repair, nevertheless it must be mentioned that the Ada Lovelace era of GeForce playing cards has been a large number, when it comes to product names and specs. Had the jolly inexperienced large caught to the GPU tiering system it used with Ampere, it would not be on this state of affairs in any respect.