For several years, developers have been asking for a native integration of Intel Quick Sync Video in their FFmpeg-based applications. We heard your requests – and are excited to announce that an Intel-supported integration is now available on the FFmpeg 2.8 release. This integration of Intel Quick Sync Video is enabling 10x faster AVC transcodes compared to Intel®-based systems not using this technology.
FFmpeg has been around for over 15 years and is probably the most popular open source media processing framework for media server applications. Intel Quick Sync Video is a hardware-based media acceleration technology available in Intel processors that integrate Intel® HD graphics, Iris™ graphics, and Iris™ Pro graphics. It can offer speed increases and improved densities for large scale media processing deployments. Until now there has not been a fully supported and optimized version of FFmpeg that enables easy access to Intel Quick Sync Video. This has made it a chore for media application developers building solutions around FFmpeg to realize the benefits of Intel’s accelerated media transcode technology. Once FFmpeg is built and installed with the free Intel® Media Server Studio – Community Edition, Intel Quick Sync Video can be enabled through a simple command line interface:
$ FFmpeg –i in.mp4 –vcodec h264_qsv out_qsv.mp4
The integration supports MPEG-2 and H.264/AVC encode/decode, VC-1 decode and, with Intel® Media Server Studio – Professional Edition, H.265/HEVC encode and decode.
The performance is pretty amazing. For example, on a single Intel® Xeon® Processor E3-1285L v4 featuring Iris™ Pro Graphics and Intel Quick Sync Video, it’s possible to transcode nine simultaneous 1080p @ 30fps AVC streams in real time at a medium quality preset. In contrast, if Intel Quick Sync Video is not used, a single processor running the same test can support slightly less than one real time stream. So – by using Intel Quick Sync Video under FFmpeg, developers will see a nearly 10X-density-increase over software only.