Intel Corporation has made a resolute step towards the improvement of performance demonstrated by ‘supercomputers’. Its new Intel Xeon Phi is called to create a real revolution in the IT world, offering unmatched performance showings on activities with the high parallelism degree. This is supposed to bring supercomputers to the new level by means of enlarging the performance levels from TFLOP to EXAFLOP. Upon the whole, Xeon Phi is the name which will be used for further line of products based on Intel MIC architecture.
Xeon Phi coprocessors are sure to enable the transmission of the reduced data via PCI slots and will primarily be applied to dedicated coprocessors. The sales start is designated to the end of the current year, and the first products presented in the market will be items which belong to Knights Corner series. According to the Intel developers’ words, Xeon Phi will firstly replenish the series of central processors Intel Xeon E5-2600/4600; further on, the company plans to integrate their new products into high-end work stations and corporate datacenters.
Xeon Phi consists of the expansion card for PCI Express wire which accommodates 22-nanometer chip with more than 50 cores and 8 GB of integrated GDDR5 memory.
The major benefits of revolutionary coprocessors are viewed in their compatibility with the tools applied for the development of software based on Intel architecture. Moreover, Xeon Phi is compatible not only with program models of x86 architecture but will also be seen as a separate computational nodes working under the control of its own OS on Linux basis. Such technological approach will provide for the flexibility pertained to the clusters creation which is unattainable with GPU solutions.