Could NVIDIA be working on a full-fat consumer implementation of the TU102 GPU? The chatter returns ahead of the holiday period.
Could NVIDIA be working on a full-fat consumer implementation of the TU102 GPU? The chatter returns ahead of the holiday period.
In July NVIDIA launched the GeForce RTX 20-series’ mid-stream refresh. Dubbed the RTX 20 SUPER series, they for the most part superseded their non-SUPER counterparts with higher shader counts and other performance tweaks. RTX 2060, 2070 and 2080 GPUs were all supplanted, but there was one glaring omission: an update to the flagship GeForce RTX 2080 Ti. Now resurfacing rumours coming from allegedly reliable sources have reignited expectation of a replacement flagship in the near future, but is that realistic?
The RTX 2080 Ti utilises a cut-down version of the TU102, NVIDIA’s biggest Turing GPU. Of the total 72 Shader Modules on board the GPU, 68 are enabled to bring the count of active shaders to 4352. A reduction in SM count also corresponds to a reduction in the memory bus width from 386-bit to 352-bit, which also accounts for the irregular amount of GDDR6 memory (11GB).
To be clear, the RTX 2080 Ti is still a mighty GPU, but not the full-fat edition that some might have expected. Until now that version has been exclusive to the $2,500 RTX TITAN, a prosumer/workstation card, but improvements in the manufacturing process or slack demand may mean that fully enabled TU102 GPUs become available for consumer markets and are thus crying out for an SKU to call their own. That SKU would be the GeForce RTX 2080 Ti SUPER.
An RTX 2080 Ti SUPER would have 72 active SMs for a total of 4608 CUDA cores, a commensurate increase in the number of RT and Tensor Cores, and a likely increase in memory bus width to 384-bit. It would probably be paired with 12GB of 16Gbps GDDR6 memory, maintaining the RTX TITAN’s position as compute workhorse with 24GB of VRAM. Doesn’t sound so far-fetched when put that way.
So there’s scope for a higher performing consumer flagship in terms of the technology NVIDIA have at their disposal, but does it make any sense from a product positioning standpoint? Enthusiasts are already grumbling over a $1000 flagship, so absent price reorganisation of the product stack (something that NVIDIA have been reluctant to do with the SUPER series) an RTX 2080 TI SUPER could have an MSRP of $1300-1500, and partner variants will be pushing $2000. That’s the sort of situation that could alienate rather than excite customers and fans.
@kopite7kimi has apparently has a high success rate with predictions, but these rumours should be viewed as uncorroborated and thus taken with a hefty pinch of salt. The RTX 2080 Ti SUPER was mooted to be on the cards as recently as August but nothing transpired, and without any competition from AMD it’s difficult to see NVIDIA’s motivation.
A launch window of January 2020 coinciding with CES has been raised, and that would be the most logical conclusion (and most convenient event) given the factors involved.
SOURCE: @kopite7kimi
