Intel announced the first beta version of its SDK for OpenCL Applications Xeon XE 2013 for Phi
A couple of days came officially parallel accelerators intel Xeon Phi 5110P , which are based in a hybrid design that combines multiple x86 cores (60 cores capable of processing 240 threads run) and 16-wide SIMD units (960 units SIMD Similar to the shader processors of GPUs).
Since the launch of its new accelerators, Intel does not rest trying to convince industry newcomer that your solution is superior to competing solutions dominate the segment servers and supercomputers (Nvidia Tesla and AMD FirePro S series), promising programming model much simpler and based on x86 .
But it should be remembered that xeon Phi is a hybrid coprocessor, and although the program its multicore x86 is very simple, you can not say the same for their Wide SIMD units 16, which require very different programming x86, and initial compliance promises Intel just released the Beta version of its SDK for OpenCL Applications XE 2013 for Xeon Phi.
The new OpenCL SDK Intel Xeon Phi allows total compatibility with code OpenCL 1.2, allowing existing applications developed with the API industry standard platform, running on their new co-processors Xeon Phi (making use of SIMD units and x86 cores) as well as in their x86 microprocessor and its IGPs Core HD Graphics 2nd and 3rd generation.
By the time the new OpenCL SDK Intel Xeon Phi is only available for Linux-based operating systems, although it is possible for a future version of Windows Server operating systems from Microsoft.
Phi Xeon 5110P will be available from January 28, 2013 at an aggressive price of $2649, much lower price than the recent Tesla K20 (U.S. $3,199, the price is still unknown K20X Tesla) and FirePro S10000 (U.S. $3599).
No doubt great news for the industry, especially for the OpenCL platform ecosystem, which is gaining more ground.
Links:2013, accelerator, application, Beta, hybrid, intel, KNC, knights corner, Linux, OpenCL, OpenCL 1.2, parallel, Phi, SDK, SIMD, XE, xeon