AMD is working on increasing the production of AMD EPYC processors and solutions based on these processors, in addition to adding TensorFlow in ROCm 1.7.

This year, the market in which AMD is succeeding is in the processor market, since its solutions are economical and offer good performance, competing with Intel in quality-price. The company is increasing production and availability of AMD EPYC processors, as they would have been in good demand. These processors are characterized by reaching up to 32 Zen cores and having 64 processing threads. One of the important, recently implemented features is TensorFlow, which is implemented within ROCm.

The company has announced an increase in the availability of solutions that have these EPYC processors as an engine, intended for the OEM market, distributors and also for integrated advanced systems such as BOXX, Hewlett Packer Enterprise, ASUS, Penguin Computing, Supermicro, Gigabyte and Tyan. The options are varied. ASUS is betting heavily on HPCs and RS720A-E9 and RS700A-E9 virtualization systems. BOXX on the other hand chooses to combine EPYC processors with Radeon Instinct acceleration for multi-GPU computing systems and for deep learning. Gigabyte directly bets on rack mount servers. Supermicro ditches its 1U, 2U and 4U systems to develop HPC and storage systems.

All this requires software and optimization, as these components are sought to give their full potential, so AMD has bet on the development of high-performance computing systems and open ROC standards, launching version 1.7. The ROC m update adds support for TensorFlow and Caffee machine learning in MIOpen libraries. TensorFlow technology is gaining prominence within the Artificial Intelligence segment and no company wants to be left out of this segment, that's why not only AMD implements it, but Google also implements it and NVIDIA does it in Volta.

Source: TPU