SIE offers high level solutions for Scientific Calculus based on Intel Phi. This card is a revolution in the mathematical coprocessors field because the x86 compatibility is full. It make Intel Phi as a serious competitor vs GPU cards that offer NVDIA and ATI.
GPU technology gives many cores and performance but its use suppose re-writing the programs, even with the help of CUDA libraries (for NVDIA).
The advantage of Intel Phi is the concept of migration “some cores” to “many cores” (old name of Intel Phi is MIC, Many Integrated Cores). This transition is very easy using the Fortran and C++ compiler by Intel. It is the first time that SIE can offer 260 cores x86 inside in one machine and in addition using all codes and programs written from 1990
For more information about the best practices:
http://www.prace-project.eu/Best-Practice-Guide-Intel-Xeon-Phi-HTML?lang=en
