Intel unveiled their new Nervana Neural Network Processors and Movidius Myriad Vision Processing Unit (VPU). Nervana chips come with two flavors, one for training and another for inference.
Aiming to conquer the artificial intelligence market Intel has been working on new, more specialized processors (namely ASICs). The company expects 3.5 billion USD in revenue in 2019 for the AI business, so it is no coincidence it releasing so many new and strong products aimed at data centers for data processing.
Intel’s Nervana neural network processors are being developed for some time by now, are being delivered to customers. Nervana Neural Network Processor has two flavors: NNP-T1000 and NNP-I1000. NNP-T1000 is for AI training, as the T in the name suggests. NNP-I1000 is the inference engine that goes through data sets to match patterns to discover rule sets.
Visual processing unit: Movidius
This kind of ASIC (application-specific integrated circuit) approach in the datacentres accelerates the many libraries and processing of data significantly, providing near-linear scalability for customers of any size. Linear scaling is not always available on each aspect of cloud applications, and energy efficiency is generally an issue. But Nervana provides one of the best-optimized solutions on the matter.
Intel has been working on another chip called Movidius, which is branded as a VPU, visual processing unit. This visual processing unit is suitable for the analysis of imagery for inference. Movidius is expected to surpass the performance of the previous generation of such ASIC incarnations by ten times. This processor is expected to be released in the first half of 2020.
Along with their hardware products, Intel also announced its new DevCloud for The Edge. Wİth Intel Distribution of OpenVINO toolkit, it will provide the developers with the latest prototype development and testing environment before they plunge into hardware expenses.