Transcript Slide 1
Graphics Processing Units (GPUs) for Powerful, Simple and Cheap parallel computing by High Performance Consulting [email protected] www.hpcsweden.se Swedish Multicore Initiative: “… nearly all commercially available microprocessors are now organized as multicore processors where performance improvements are obtained through parallelism.” Parallel programming is necessary “… only 1% of software developers are proficient in parallel programming” We make parallel programming simple “parallel software development is 2-3 times more expensive than conventional software development.” We make parallel programming cheap GPUs have 35 times the performance of CPUs plus a simple parallel programming model GPUs are cheap. They use a simple programming model yet provide enormous levels of performance in an energy-efficient manner. Moreover, the simplicity of their programming model means that code written today will scale linearly with future hardware performance. GPUs are used in: •Image analysis •Signal processing •Finance •Computational Fluid Dynamics •Medical imaging •Rendering •Simulations and real-time •Plus much more! “If we think about how long it would take to handle this much complexity with traditional methods, we’re probably close to 100 times faster.” - Weta Digital, Visual effects Vendor of Avatar “With parallel processing on GPUs, pricing a large portfolio of exotic contracts can be accomplished in minutes instead of hours.” - Curt Randall, Executive Vice President of SciComp “GPU computing technology has given us a 100-fold increase in some of our programs, and this is on desktop machines where previously we would have had to run these calculations to a cluster.” - John Stone, senior research programmer at the University of Illinois “By using … GPUs, these applications can now be run 10-20 times faster, which means even a PC with Tesla GPUs can outperform a supercomputer." - HPCwire High Performance Consulting Consultants in GPU programming and computations. [email protected] www.hpcsweden.se