Performance evaluation of heterogeneous computing systems based on modern IBM POWER processors

The article is devoted to the complex study of hardware and software of heterogeneous computing systems based on modern IBM POWER processors and NVIDIA Tesla graphics coprocessors. Using the various parallel programming technologies, the performance of the memory subsystem and central processors is investigated in parallel mode. The effectiveness of the functioning of math libraries has been studied, including those providing offloading calculations to the coprocessor. Basic recommendations on the use of this class equipment for solving various scientific problems are given, based on the results of the work carried out.