Intel in recent months has introduced two specialized Nervana platform processors: NNP-I and NNP-T. The first was shown in June, and the second in August.
Today Intel decided to introduce new products again, but more specifically. So, in fact, the solutions are called NNP-I1000 and NNP-T1000. Both are focused on working with artificial intelligence, but the first is designed to handle logical inference tasks, and the second is for deep learning.
In addition, Intel introduced a new generation of Movidius VPU processor, which will be available to customers early next year. As for the rest of the new products, Intel notes that they are in the process of production and delivery to customers.
Intel in the press release says nothing about the parameters of the new solutions, but this was known back in the summer. Recall that the NNP-I1000 is a specialized modification of the 10-nanometer Ice Lake processor, which has been removed from “everything superfluous”, including the GPU, but various blocks for working with AI have been added.
NNP-T1000 is a very large specialized processor with 24 tensor cores and performance up to 119 TOPS.