Intel prepares an armada of Phi Xeon cGPUs end of this year
Last week intel announced its change of strategy for their cGPUs formerly known as Knights Corner, which are now known as Xeon Phi , a change that not only reflected in its title business, but its shift in focus from direct contracts with manufacturers of supercomputers, to the mass market.
Phi xeon will not be a single product but a complete fleet of products destined to conquer the market for high performance computing (HPC) as displacing cGPUs Nvidia (Tesla) and AMD (FireStream) , and will be available in multiple editions various configurations with more than 50 x86 cores / SIMD vector units with frequencies above 1 GHz, with up to 16GB of ECC memory and computing power than TeraFlop in double precision (FP64).
It is assumed that even the cGPU Nvidia Tesla-based K20 core 2 nd generation Kepler GK110 (1 to 1.3TFlops FP64) will be able to compete against increasingly powerful Xeon Phi, and also the ease of porting the code Phi Xeon applications to work a few days will be compared with the several months it would take this work in a traditional cGPU will be the turning point that will break down the balance in favor of the Intel solution.
Link: Intel Xeon Phi steals top HPC tenders, more SKUs to come (VR-Zone)Tags: 2012, cGPU, coprocessor, GPGPU, HPC, intel, Knigths Corner, Larrabee, mic, navy, Phi, SKUs, SMP-on-chip, xeon