GPU Computing & Accelerators
There is an ever increasing demand of computational power in scientific or industrial research. At the same time, high-performance computing systems consume very much power. The deployment of accelerator technology for additional and specialized computing hardware components constitutes a big leap in computational power while, at the same time, lowering energy-consumption significantly.
For building future Exascale high-performance computing systems, a low energy consumption is key. Even many of the current TOP500 list supercomputing systems possess accelerators as a core ingredient, thus achieving their excellent performance.
Graphics Processing Units (GPUs) as well as specialized x86-based many-core systems may act as accelerators. These co-processors are able to perform certain parts of the scientific calculation more efficient. Tasks that are computationally intense and highly parallelized are not executed on the CPU, the system’s main processor, but are transferred to the co-processor. This way not only a significant reduction of time-to-result is gained, but also the energy consumed is lowered.
GPUs - Graphics Processing Units
Graphics processors (GPUs) provide a significant acceleration of scientific applications in the areas of life sciences, computational fluid dynamics (CFD), bioinformatics, machine learning, quantum chemistry, and material sciences, to name just a few.
CPUs - Central Processing Units
CPUs consist of only a few execution units which are optimized for the serial processing of data. They can do a variety of calculation types. GPUs and other kinds of co-processors on the other hand possess a large number of small but highly efficient computation cores, highly optimized for a certain type of computational task, as a whole constituting a massively parallel architecture. The computational results are then transferred back to the CPU, where they are integrated in the total calculation.
Many-core co-processors based on x86 architecture allow for an easy reuse of existing codes, with an increase of the package density as a side-effect. Standard tasks that normally run on an x86 processor, are now executed in a highly parallel way on many cores inside these accelerators.
transtec offers a wide range of accelerators. We are happy to advise you on the choice for the accelerator technology that fits best. We will provide you with a perfectly optimized total solution that accelerates your applications significantly.