Despite the coronavirus pandemic, in general, the IT market still continues to exist without much change. New devices are still announced and go on sale, except that various exhibitions and events are canceled.
However, the output of a new generation of Nvidia video cards, according to the source, has been moved, at best, to the fourth quarter of this year. True, the reason for this was not only the coronavirus.
An insider claims that Nvidia wants to evaluate the capabilities of the upcoming “big Navi” – AMD’s new GPU with RDNA2 architecture, which will serve as the basis for the long-awaited top-end Radeon graphics cards. Moreover, now the flagships of GeForce simply have no competitors – Nvidia may well wait.
As for AMD’s new products, there is no clear data about them yet, but we can understand something from the presentation of the Xbox Series X. Recall that the platform of this console contains a GPU with 52 CUs, which means 3328 stream processors. In comparison, the Navi 10 has 40 CU units. That is, at similar frequencies, the performance of a video card with 52 CUs should ideally be about 30% higher than that of the Radeon RX 5700 XT, and this without taking into account architectural improvements, and they will. A 30 percent increase over the RX 5700 XT is already the level of the RTX 2080 Super.
Given the fact that, with a high degree of probability, the Big Navi will contain much more computing units than the Microsoft GPUs, you can count on very impressive performance that significantly exceeds the capabilities of the RTX 2080 Ti.
True, the conditional RTX 3080 Ti will also obviously be much faster than the current flagship, so in this matter even comparing absent cards in absentia is pointless.
Returning to the Ampere cards, it is worth starting to recall the recent rumors about their parameters. The new data almost completely copies that information, but with one important correction: previously it was believed that the RTX 3080 Ti will receive a GA102 GPU with 5376 CUDA cores (the maximum number of cores in a GPU, the card itself may have fewer ones), and the RTX 3080 – GPU GA103 with 3840 CUDA cores, now the source says the RTX 3080 Ti will get the GA103 with 3840 CUDA cores.
At the same time, there is a chance that in reality there will be twice as many CUDA cores if their number in each SM block is doubled. In addition, the source talks about doubling the RT cores, that is, when effects with ray tracing are activated, performance will sag much less.