BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20141119T220000Z
DTEND:20141119T223000Z
LOCATION:Show Floor - 233 Theater
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Current trends in processor technology have led to bottlenecks in memory latency, data movement, and storage. A need exists to accelerate practical exascale computing by focusing on power efficiency, simplicity of component design, and hierarchically designed systems. The only way to build a truly high performance computing system capable of meeting exascale demands and beyond is to design the simplest and thus most efficient processing unit, without sacrificing functionality or scalability.=0A=0ASystem Fabric Works (SFW) and REX Computing have been jointly developing a full node architecture centered around a new processor approach that is optimized for HPC.  We propose to build a grid of many (potentially thousands of) Multiple Instruction, Multiple Data (MIMD) cores to execute instructions independently, drastically reducing the demands on memory bandwidth and thus greatly increasing the efficiency at which typical HPC workloads may be performed.
SUMMARY:A New Node Design for HPC
PRIORITY:3
END:VEVENT
END:VCALENDAR