The news will surpass those originally thought, at least according to sources. The more details we have already updated in the past weeks and months, the more we seem to be familiar with.
NVIDIA RTX 3080 Ti, full of news even on its heatsink
The first thing that strikes us is that it looks like NVIDIA will make things difficult for AIB with its new heatsink, though it looks like it will have a pretty cool design similar to the RTX 2000, this time it will involve three fans instead of two.
Another detail is the possibility that at least the top three cards contain two of the eight connectors. The reason for the appearance is that NVIDIA will try to capture higher clocks than the other series depending on the chip, which will require more power.
As for the ports to install, sources say three DisplayPort 2.0, HDMI 2.1 and USB-C, so these new graphics cards will be compatible with the new serial levels. The other good news is that they will all add up PCIe 4.0 x16 to his credit, thus giving a step to the new generation of this bus as AMD has already done.
Rumors point back to the 5376 Shaders setup
GA102 will be NVIDIA's new TOP TOP chip. As we have already said at the time, it said that a chip in its grand plan would have 5376 CUDA Coresso GA100 should exceed that value in HPC.
The reduction of the lithographic process allowed this fact, where in addition, it would seem, the use would not increase and would somehow decrease, being between 220 watts and 230 watts. The RTX 3080 Ti can reach with these figures the speed is around 2.2 GHz
Another new feature we've introduced in the last few weeks is the memory installation at 18 Gbps, which can provide bandwidth for 864 GB / s (+ 40% compared to the RTX 2080 Ti). To be more specific, NVIDIA can squeeze the tide down 2.5 GHz on the GA103 and GA104 chip in the best cases, though it is not clear whether this data has been excessively removed by the manufacturer or stock details.
This increase in power results in the least + 40% restructuring compare the RTX 2080 Ti to 4K, in the worst case. So much so, that the middle grades will be able to stream games in 4K as the RTX 2080 Ti is now reaching, so we're talking a huge jump in performance.
Lots of great performance on Ray Tracing, Tensor Cores double and no loss of performance
Obviously, the number of RT Cores did not increase significantly, but they would achieve so much performance that if they were implemented in the Load, they would be able to perform intervention procedures 4 times faster.
What will increase the value of Tensor Cores is to double, which together with the development of RT Cores will make Ampere's lower range of the same Ray Tracing and DLSS as currently high in Turing.
If all that is ripe is true, we're talking about a huge step forward for NVIDIA that will improve Ray Tracing and DLSS, just as they are dense. Testing will be given a larger backdrop than we had between Maxwell and Z arrange for himself, while also seeing a surprisingly high performance that can cost NVIDIA a set, as long as it's not the same as the current RTX 2080 You and I larger numbers.