The development of the new generation of video cards continues unabated, between rumors and leaks of all kinds. Last week rumors about a possible high-end AMD GPU went around the world but turned out to be unfounded. This time, however, the discussion is different, we are not dealing with an image stolen from who knows which Asian fair, but with a real benchmark, with lots of images that testify to the goodness of this leak.
The two leaked NVIDIA video cards are characterized by a significantly higher number of CUDA cores than the current models, however, it is not known if they belong to the next consumer lineup or if they are professional GPUs, the certainty is that these tests demonstrate the existence of these graphics chips. The much rumored net increase in performance of the RTX 3000 may therefore not be just a mirage.
The power of 7 nm
NVIDIA manufactures its video cards with a 12nm manufacturing process, against AMD's 7nm. The next generation of GPUs will instead be built at 7 nm. Among the benefits of refining the production process we find one more space in which to insert the calculation units. NVIDIA therefore has almost double the usable surface available compared to the current Turing video cards so, at least theoretically, the recently leaked benchmarks and features are more than plausible.
These are two high-end products, of which the most powerful has scored a score of 184.096 on GeekBench. For comparison, an RTX 2080 Ti reaches around 130,000 points in this test, for a gap that settles around 30%. Be careful though, this number could also go up, because the clock of the GPU used in the test is 1.1 GHz, lower than that of the RTX 2080 Ti. However, the frequencies may be lower to compensate for consumption and the heat generated by such a large number of Cuda Cores, 7552 in all (118 Compute Units), against 4352 of the RTX 2080 Ti.
The second GPU leaked online instead takes advantage of 6912 Cuda Core (108 Compute Unit) and reached a score of 141,654, which however remains higher than that of the RTX 2080 Ti. The data relating to memory is also interesting. The most powerful GPU mounts in fact 48 GB of RAM, the second one has 24, a detail that suggests that they are video cards of the Quadro range, belonging to the professional series of NVIDIA.
Despite this, the interest in this leak remains high even in the consumer sector, because at the Cuda Core level the differences between the Quadro and RTX ranges are minimal, just think that the current top of the Quadro 8000 range has only 25 Cuda Core more than the RTX 2080 Ti. This suggests that the next generation of NVIDIA video cards for gaming could focus on a much higher number of Cuda Core compared to a lower clock, a choice that seems plausible even if all confirmed in the coming months.
