By 2018, Silicon Graphics International (SGI) plans to build supercomputers that are 500 times faster than the most powerful systems today, using Intel's many integrated cores (MIC) architecture.
Chips based on the MIC architecture combine standard x86 cores with specialized cores to improve high-performance computing. "[MIC] gives us the compute density we need" to support exaflops of performance by 2018, says SGI's Eng Lim Goh.
The MIC architecture solves the problem of the time and cost required to produce the proper accelerators by including many specialized cores in a chip able to run standard x86 software. Intel recently demonstrated its first experimental MIC chip, which has 32 cores and combines vector processing units with standard central processing unit cores. A Xeon server with eight such chips can deliver 7.4 teraflops of performance, according to Intel's John Hengeveld.
Intel also is developing the software ecosystem before it releases the first commercial MIC chip, as well as backing the development of OpenCL, a parallel programming framework. "Our intent is to have OpenCL support available on the first MIC product--Knights Corner," says Intel's Radoslaw Walczyk.
From IDG News Service
View Full Article
No entries found