Skip to Main Content
We investigate the scalability, architectural requirements, and performance characteristics of high performance systems. We use the BlueGene/L (BG/L) project as a case study for deep computing applications. This massively parallel system of 65,536 nodes with autonomic features is based on a new architecture that exploits system-on-a-chip (SoC) technology to deliver a target peak processing power of 360 teraFLOPS (trillion floating-point operations per second). BG/L is operational at a price/performance and a power consumption/performance targets unobtainable with published conventional architectures.