GPU Computing

GPU computing is something which is fast becoming a very important feature on "video" cards. NVIDIA have a number of useful utilities for CUDA enabled cards which all help to enhance the performance of day to day tasks. On the AMD front we have "Stream" applications which use the GPY to enhance performance. Here we have taken a few examples of the different technologies at work to find out how the cards benefit us in real world use.

Windows Drag and Drop Transfer
The cards are able to take advantage of GPU acceleration for transfers to compatible media players. This means that when we copy a video file to a player Windows recognises if the format is compatible or not. If a conversion is needed we are asked if we would like to convert and copy as one task and should we choose yes the system uses our GPU to enhance performance. This makes the task more simple than converting and copying as two separate tasks, quicker and often cheaper as we don't need to buy a 3rd party utility to convert the files.

Our test file is a 42 minute AVI file.


GPU Assisted Media Conversion
Our second media encoding test takes a 4GB 1920x1080 file and converts it with Arcsoft Media Convertor 7 to iPad format. One run is completed with CPU only encoding, the second and third with ATI Stream /NVIDIA CUDA (GPU assistance).


When looking at the Windows Drag and Drop Transfer results there are very clear advantages to be seen from using our GPU to assist the CPU. In the case of our media conversion test the results look less impressive but what they don't show is that the GPU assisted results were achieved with less CPU use. The PCS++ was able to free up around 8% of CPU performance for other tasks while the encoding was taking place.