GPU computing - high performance computing at its best

A graphics processing unit (GPU) is a RISC processor used in video game systems and computers for calculating the graphical output for monitors. GPUs are used at manipulating and displaying computer graphics. GPUs are a part of mother board or they can come along with video cards. Graphic accelerators are a part of GPU also. High performance GPU computing involves solving advanced computational problems using supercomputers and computer clusters. GPU computing includes all tasks that a graphics processing unit can calculate.

In GPU computing the CPU calculations are replaced by Graphics Processing Units. Migrating large scale algorithms and entire kernel onto the GPU co-processors help in arriving at the answer much faster and thus decreases the processing time. GPUs are never a complete replacement for CPUs but complementary. Parallel operation of CPU and GPU has found to increase the performance. CPU’s offload the tasks which are better performed by GPU leading to high performance computing. GPU excel CPU’s in certain computational tasks.

A GPU’s processing power is mostly used while dealing with 3D scenes and video. The GPU computing for a graphics task such as rendering a 3D scene involves many stages like vertex processing, vertices grouping, fragment processing and mathematical operations. GPU computing is also known as General Purpose Computation on Graphics Processing Unit. Acceleration of video and picture editing, artificial intelligence and physics calculations come under the tasks of GPU computing.

GPU uses Advanced Graphics Port (AGP) to communicate with the mother board of a computer. Programmable shaders capable of manipulating vertices and textures are supported by latest GPUs. Very high precision color spaces, over sampling and interpolation techniques to reduce aliasing are also supported by GPU. GPU can process multiple independent vertices and fragments in parallel. Moreover a variety of computational resources are available on the GPU.

Multi-GPU’s are also available and are generally used in high-end home computers to accelerate computer games. Multi-GPU systems use more than one GPU. An example of multi-GPU system is Nvidia which can make a PC perform 250 times faster than usual. Not only that but by making use of the CUDA architecture which is made up ofhundreds of processors cores data set can be easily crunched via the application. General Purpose computing on graphics processing unit uses a GPU to perform computation in applications traditionally handled by CPUs. Additional programmable stages are included to make this happen. The general Purpose Computing on GPU(GPGPU)has cleared new ways in diverse fields like scientific image processing, oil exploration, linear algebra and 3D reconstruction.

GP GPUs are used in computer clusters, physics engines, FFT, audio signal processing, digital image processing, bioinformatics, scientific computing, DSP, neural networks and many more. GPU computing CUDA was developed by Nvidia. AMD has announced a super computer based on GPU which will be built of more than thousand GPUs. This supercomputer will have a computing power of one petaflop. The processing power of GPUs have further increased due to the advent of OpenGL API and similar functionalities in DirectX. Intel, AMD/ATI, NVIDIA and S3 Graphics are some of the players in the market who produce GPUs. Many companies are now engaged in researches to find out the further benefits of GPU computing.