Hardware encoding is more pervasive with Intel’s Quick Sync embedding CUDA GPUs inside GPUs plus NVIDIA GPUs have MPEG NVENC encoding support so how does it compare with software encoding? For HEVC, can Xilinx’s FPGA solution be a boost in terms of quality or cost compared to software encoding?
Jan Ozer has stepped up to the plate to put this all to the test analysing how many real-time encodes are possible on various cloud computing instances, the cost implications and the quality of the output. Jan’s analytical and systematic approach brings us data rather than anecdotes giving confidence in the outcomes and the ability to test it for yourself.
Over and above these elements, Jan also looks at the bit rate stability of the encodes which can be important for systems which are sensitive to variations such services running at scale. We see that the hardware AVC solutions perform better than x264.
Jan takes us through the way he set up these tests whilst sharing the relevant ffmpeg commands. Finally he shares BD plots and example images which exemplify the differences between the codecs.
FPGAs are flexible, reprogrammable chips which can do certain tasks faster than CPUs, for example, video encoding and other data-intensive tasks. Once the domain of expensive hardware broadcast appliances, FPGAs are now available in the cloud allowing for cheaper, more flexible encoding.
In fact, according to NGCodec founder Oliver Gunasekara, video transcoding makes up a large percentage of cloud work loads and this increasing year on year. The demand for more video and the demand for more efficiently-compressed video both push up the encoding requirements. HEVC and AV1 both need much more encoding power than AVC, but the reduced bitrate can be worth it as long as the transcoding is quick enough and the right cost.
Oliver looks at the likely future adoption of new codecs is likely to playout which will directly feed into the quality of experience: start-up time, visual quality, buffering are all helped by reduced bitrate requirements.
It’s worth looking at the differences and benefits of CPUs, FPGAs and ASICs. The talk examines the CPU-time needed to encode HEVC showing the difficulty in getting real-time frame rates and the downsides of software encoding. It may not be a surprise that NGCodec was acquired by FPGA manufacturer Xilinx earlier in 2019. Oliver shows us the roadmap, as of June 2019, of the codecs, VQ iterations and encoding densities planned.
The talk finishes with a variety of questions like the applicability of Machine Learning on encoding such as scene detection and upscaling algorithms, the applicability of C++ to Verilog conversion, the need for a CPU for supporting tasks.
Delivering great quality, live video without breaking the bank is difficult. This talk looks at the different ways companies are dealing with this challenge.
NGCodec’s founder, Oliver Gunasekara, starts by quantifying the millions of dollars spent just by one company each year just on delivering their video and introduces the difficulties of CPU encoding compared to dedicated chips – ASICS and looks at how FPGAs fit in. Cloud-based FPGAs are available on AWS, Baidu, Alibaba and others.
After covering Twitch’s move to VP9 on FPGA, the talk finishes looking at on-premise implementation, Oliver looks at the cost of ownership of servers compared to Xilinx FPGA.
Founder. & CEO,
Subscribe to get daily updates
Views and opinions expressed on this website are those of the author(s) and do not necessarily reflect those of SMPTE or SMPTE Members.
This website is presented for informational purposes only. Any reference to specific companies, products or services does not represent promotion, recommendation, or endorsement by SMPTE