They parallel accelerators Intel Xeon Phi
Announced five months ago , today finally made his debut the new intel Xeon Phi parallel accelerators “Knights Corner,” Intel’s first product with facing solutions for accelerated parallel computing FirePro and Nvidia Tesla S Series (formerly known as FireStream) of AMD.
Intel promised performance in double precision calculations over one Teraflop, ending just fulfilling promise, offering a yield of 1.07 Teraflops on the pinnacle of its new family of parallel accelerators: Xeon SE10P/SE10X Phi. The specifications of these new accelerators (cGPUs) are:
Xeon and Xeon Phi Phi SE10P versions SE10X are “tailored” (61 cores) developed for the Stampede supercomputer (delivered almost a year ago), so its TDP is considerably higher than that of the new Xeon Phi 5110P (60 nuclei), which Intel calls “the final version” of his cGPU Knights Corner. Also expected future Phi Xeon 3100 Series (57 cores), which will be available from January to June next year.
Maybe pure double precision performance, the new Xeon Phi not sound as impressive compared to 31.1 Teraflops K20X Tesla and 1.48 Teraflops of FirePro S10000 , but it compensates Intel x86 offering performance nearly triple to 25 Intel Xeon CPUs working together in parallel x86 intensive applications, and offering a lower consumption of three (Xeon CPUs), a factor that should pique the interest of many companies.
Phi Xeon 5110P is now available for OEM market (OEMs) and will be available to consumers on January 28, 2013.
Link: Intel joins the data-parallel computing with Xeon Phi fraternity (The Tech Report)Tags: 3100, 5110P, cGPU, computing, coprocessor, intel, KNC, knights corner, Larrabee, mic, parallel, SE10P, SE10X, x86, Xeon Pri