Computing in the Many-Core Era
Many-core processors provide compelling performance and performance per watt solutions for many technical computing problems. Utilizing parallel programming techniques, computing work can be broken up into many light-weight threads that can run concurrently on a chip, getting a lot of processing done during each chip cycle. These processors are commonly known as coprocessors, because they need to work with a standard microprocessor that runs the operating system, communicates across the network, etc. Various many-core options exist in the market today including FPGA's, GPUs, and the new Intel® Xeon Phi™ coprocessor.
SGI, a Leader in Coprocessing
SGI leads the industry in delivering many-core coprocessing solutions, dating back to the Geometry Engine™ which accelerated graphics applications in the 1980s. SGI then co-developed the SGI Tensor Processing Unit (TPU), followed by RASC™ technology, FPGA's that were tightly-coupled to our shared memory architecture. With RASC technology, SGI created the world's largest single system image server with accelerators, to solve the most challenging life-sciences problems. With a full team of application experts, SGI has a unique position to help customers solve problems with many computing technology and has services and support personnel ready to help customers port and debug specific applications. Solutions are integrated with SGI software and SGI InfiniteStorage™ to provide complete solutions for customer workflows. Both scale-up and scale-out coprocessing solutions are available to tackle any type of problem in scientific research, product development and homeland security.