Comparison of General Purpose Graphic Processor Units as a Substitution for Traditional Processors

This paper describes two leading general purpose GPU (Graphics Processing Unit) computing technologies, ATI Stream and NVIDIA CUDA (Compute Unified Device Architecture) platforms which use GPU as parallel compute device. Programming tools, CUDA C and OpenCL (Open computer Language) allow to demonstrate the usage of the GPU as a parallel computing device on a simple vector multiplication algorithm. Results of several experiments, executed by NVIDIA Geforce 9800 GT and ATI Radeon HD 4650 GPU, show that NVIDIA gives better results. It is necessary to take into account that GPGPU is still new and evolving concept, as well as the fact that looking at specifications and price of both graphics cards, Geforce 9800 GT GPU is better than ATI Radeon HD 4650, in terms of GPGPU computing. GPGPU graphics cards are important in scientific calculations, in computationally – intensive applications, because this kind of applications can use all of their parallel power and acceleration.