There is no doubt that AMD has given NVIDIA where it hurts the most and although Huang’s had been warned for months, the movements are arriving late and badly, especially for current owners of an Ampere card, who will see how their models are quickly left behind. To compete against the RX 6700 and RX 6800 NVIDIA plans to launch the RTX 3070 Ti 10 GB, which has just confirmed that it will have 10 GB of VRAM with GDDR6X !
Be it late or not (something buyers will decide) NVIDIA is not going to give up so easily, especially with its current potential in the market. On the contrary, it pulls out muscle as it already did with RDNA 1.0 and the RX 5000 presenting its SUPER range, except that in this case and given the close launch of its Ampere GPUs, the movement will be to launch Ti versions to compensate for the shortcomings of its current models.
Is an RTX 3070 Ti necessary in the market with the RTX 3070 and RTX 3080?
Well, having seen the latest engines and games, the truth is that it will be a palliative for the greens, and yes, it is a necessary model since the RX 6800 has taken the shame out of the two existing models in NVIDIA.
Recall that the RTX 3070 gets 8 GB GDDR6 VRAM configuration with a 256-bit bus, while the RTX 3080 achieves 10 GB GDDR6X with a 320-bit bus. Its direct rival, the RX 6800, has been endowed with 16 GB of GDDR6 with a 256-bit bus plus the Infinity Caché that acts as an L3 with 128 MB, thus making up for the bus difference compared to its superior rivals.
What is NVIDIA going to do to try to better position itself in the market? Well, very simple, you plan to launch your RTX 3070 Ti with 10 GB of GDDR6X with a 320-bit bus, or what is the same, you plan to embed a GPU that is basically a decaffeinated RTX 3080 that competes with both AMD models.
It is not yet clear, but rumors and leaks indicate that we will be facing a GPU that is closer in performance to the RTX 3080 than to the RTX 3070 Ti. There are two possibilities where one of them by sheer logic is plausible, or at least more likely:
- NVIDIA will use the GA102-150-KD-A1.
- It will develop a new chip that has been around since the beginning: GA103-300-A1.
Which one is more likely to see the light? The GA102-150-KD-A1, mainly because it would be easier to place the defective chips of the RTX 3080 in this new RTX 3070 Ti, since we went from 68 SM to the more than probable 58 SM of this new model.
This would imply a total FP32 Shader count of 7424 , or what is equal, 3712 real Shaders attached to a 320-bit bus to give life to those 10 GB of GDDR6X that are giving such good results to the current RTX 3080.
Of course this will be another critical GPU as users expect more VRAM with a view to changing their graphics card in a few years, but for that NVIDIA is shaping the RTX 3080 20GB (presumably), so it would cover these users who normally play in 4K at high FPS.
In any case, this RTX 3070 Ti 10 GB should arrive with a new PCB called PG132-SKU40 , which has also been confirmed by the leakers, so it seems to be closer to its release than we think.
Source :- kopite7kimi