ATI Stream: Using GPU Resources for Calculations
Another feature of the integrated AMD 785G chipset is the support of ATI Stream technology that allows using Radeon HD 4200 GPU resources for computational tasks.
However, despite all AMD’s efforts aimed at promoting this technology, we don’t consider it a serious advantage in favor of integrated AMD solution. The thing is that at this point its software support is still at a very early stage that is why so far we can only speak of its potential and not of its real benefit for the end-users.
To demonstrate the way Stream technology works, AMD use video transcoding from one format into another using Cyberlink MediaShow Espresso utility. And in this case everything does really work. There appears a check mark next to “Enable Hardware Decoding” in the utility settings on AMD 785G based board. This option enables transcoding acceleration using shaders of the integrated Radeon HD 4200 graphics processor.
Stream activation really speeds up the transcoding process. As an example, we measured the time it took to transcode a two-minute video initially recorded in H.264 720p format into appropriate format for YouTube.com. The results turned out very impressive:
As we see, using graphics core shaders for video transcoding allows to speed things up by almost 40%. At the same time the CPU utilization gets lower, which makes the system more responsive to user actions even when Cyberlink MediaShow Espresso is running.
And everything could be considered great if it weren’t for one thing. At this point the use of GPU resources for video transcoding instead of the CPU not only speeds up the processing, but at the same time lowers the quality of the final videos. You can see what I am talking about from the screenshots below:
So, at this time ATI Stream support in AMD785G is of pure theoretical interest to us and can hardly be considered a serious technological advantage of this platform.