Nvidia CUDA

By | October 15th, 2009



Nvidia Cuda™ (Compute Unified Device Architecture) is a parallel programming architecture that enables to program and run highly parallel and compute extensive tasks on you GPU.

nvidia cuda Nvidia CUDA

In 2006 Nvidia announced it’s CUDA architecture supporting C language to program the programmable Graphics Processing Units. Before that only games could run on the GPU, but with CUDA you can execute any code on the graphics card. As we know a graphics card offers a lot of streaming multi-cores capable of performing tasks in parallel hence provides a great resource for parallel computing.

 Nvidia CUDA

This figure explains it all, the G92 GPU which is present on famous 9800 GTX graphics card can give a massive performance as compared to a Core 2 Duo processor. Imagine you have all the milticores of GPU in your hand and free to use them in any desirable way. This changes everything, many existing algorithms and applications have been converted to CUDA code and showed 100 times or more increase in the performance. It is like how you find Broadband after years of slow dial-up systems– the difference is striking! Many who try the G92 GPU never go back to their old graphics processor. A must try, indeed.

Learn Nvidia CUDA

Learning CUDA is very easy for programmers as the native language for CUDA is C. Nvidia website has tons of tutorials and guides to get you started with GPU programming. I’ll mention the important links here:

CUDA_Programming_Guide.pdf.

Tutorials and Lectures:

For comprehensive tutorials and lectures on this technology visit this link.