The announcement of the Ampere architecture is still set for the May 14 at GTC 2020, but NVIDIA will not present the RTX 3080 Ti until the end of this year. We still do not have an official date, but the most recent information points to September or October, which would leave us with a launch for the end of the year. If everything goes according to plan, the graphics card should come with the RTX 3080.
Last week we shared with you a good dose of information about the 30 series RTX that, in general, remains, although it has received an important update with some changes and news It is worth highlighting and dealing with in a new article. The first thing that we must bear in mind is that the forecast of improvement in the CPI is reduced, which goes from between 10% and 20% to a fixed 10%.
Despite that negative change, the forecasts for the huge increase in ray tracing performance remain. This means that Ampere will multiply by four the Turing performance in ray tracingYes, but what about rasterization performance? According to this new information, the RTX 3080 Ti will be up to 70% more powerful than the RTX 2080 Ti (in the best possible scenario), and up to 40% more powerful than that in the worst possible scenario.
Possible Specifications of the RTX 3080 Ti
336 texturing units.
96 raster units.
168 RT cores.
12 GB of GDDR6 at 18 GHz.
864 GB / s bandwidth.
Up to 2.2 GHz frequency on the GPU.
21 power TFLOPs.
Two 8-pin power connectors, with a TDP of 230-250 watts.
Compatible with PCI Gen4.
The RTX 3080 Ti will feature a new cooling system based on three fans and will have a simplified construction that will seek to reduce costs without affecting quality. As planned it will come manufactured in 7nm process, although it is commented that the mid-range models of the Ampere series will use the 8nm manufacturing process.
In terms of performance, this new filtration highlights that Ampere will no longer have to choose between one of the three improvements that ray tracing offers: global illumination, shadows and reflections, since it will have enough power to work with ray tracing applied simultaneously to these three technologies, and its impact on performance will be less compared to what we have seen in Turing.
It may seem exaggerated, but if Ampere really multiplies Turing’s ray tracing performance by four it would be perfectly viable. On this matter it is important to remember that NVIDIA will, in theory, double the ratio of tensor cores for each SM unit, which in the case of the RTX 3080 Ti would leave us with a total of 168 RT cores. To give you an idea of what this means, I remind you that the RTX 2080 Ti has 68 RT cores.
We will be attentive to the online event that NVIDIA will offer on the occasion of the GTC 2020 and we will tell you all the key news about Ampere, an architecture that, we remember, will compete with AMD’s Radeon RDNA 2 series, which as we know has been used on new generation consoles and whose debut is scheduled for later this year.