Intel, AMD and NVIDIA 3 potentiate the world’s most efficient supercomputers
Launched just three days, parallel accelerators (cGPUs) intel Xeon Phi 5110P , amd FirePro S10000 and Nvidia Tesla K20X get the top three among the 500 supercomputers with higher performance per watt in the industry.
The world’s most efficient supercomputer is beacon (Appro GreenBlade GB824M) of the National Institute for Computational Sciences, University of Tennessee, consisting of microprocessors Intel xeon E5-2670 and Intel Xeon Phi 5110P accelerators. Its power is gigaflops per watt 2499.
The second place is held by the supercomputer SANAM (Adtech ESC4000/FDR C2) of King Abdulaziz City for Science and Technology, Intel Xeon microprocessors composed E5-2650 and AMD FirePro accelerators S10000. His power is of 2,351 gigaflops per watt.
Finally we have the Titan supercomputer (Cray XK7) of Oak Ridge National Laboratory, consists of AMD Opteron microprocessors and accelerators Nvidia Tesla 6274 K20X. Its power is gigaflops per watt 2142.
Among the first 10 locations have in 4th place supercomputer Todi (Cray XK7), formed by microprocessors AMD Opteron 6272 and accelerators Nvidia Tesla K20, the 5th to 10th place consists of supercomputers based on IBM BlueGene / Q (CPU IBM Power BQC) .
No doubt good news for manufacturers of parallel accelerators, whose multiple SIMD units (shader processors) have a massive computing power, and allowed them to capture the first locations.
Special mention deserves the supercomputer Titan (2nd place in the Green500), which in turn is leading the top500 supercomputer among the most powerful supercomputers (though more powerful does not mean more efficient).
Link: Heterogeneous Systems Re-Claim Dominance Green500 List (Green500)Tags: amd, beacon, efficiency, FirePro, Green500, intel, K20X, Nvidia, performance, Phi 5110P, S10000, SANAM, server, supercomputer, Tesla, Titan, top500, watt, xeon