HPCwire

Lawrence Berkeley National Laboratory's Horst Simon says in an interview that he doubts exascale computing will be reached by 2020 because "the methods we have been using to predict when we cross certain thresholds, like teraflop/s and petaflop/s no longer apply." Simon notes that it remains unclear what architecture will bring computing to the exascale level, and he says straight line extrapolation thinking is flawed and heading for a major disruption due to a fundamental technology transition. Simon says exascale architecture development is following three branches: a multicore path centered around high-end central processing units; a Manycore/embedded approach that employs simpler, low power cores from embedded systems, and the graphical processing unit (GPU)/accelerator approach using specialized processors from the gaming/graphics market sector. To read further, please visit http://www.hpcwire.com/hpcwire/2013-05-15/no_exascale_for_you_an_interview_with_nersc_s_horst_simon.html.