最环保的编解码器:LCEVC

One of the most fascinating recent threads on LinkedIn focuses on the need for video compression technology to "go green." It's a discussion that all of us in the streaming video space should be thinking about.

这个帖子是由David Ronca发起的, current director of video encoding at Meta and former director of encoding technologies at Netflix. 12月. 8, he 发布, “基于目前的趋势, global data centers will consume > 20% of global power in 2025 (from ~3% in 2020). The primary drivers of this growth are video processing and video understanding. ... We do not need new generations of encoders that increase global compute (and pollution) by an order of magnitude. 我们需要解决污染问题的编解码器."

The responses and suggested solutions to this issue were varied and complex. Fabio Sonnati, media architect and encoding and streaming specialist at 日本电报电话公司的数据, 主张我们仍然需要更先进的编解码器, 声明, “我想是长期的, 中短不太确定. 除了8K, AR/VR是否会起飞, other data compression problems will substitute video and the sustainability challenge will remain." Ronca pushed back, 声明 that "pipes continue to widen by as much as 50% per year. If you have 100Mbps how valuable is reducing video bitrate from 3.5Mbps到2.7 mbps?"

首席运营官Alex Liu NETINT技术, proposed that "one solution to this paradox is using domain-specific architecture ASIC to improve compute efficiency. 与软件相比， ... 专用asu (VPU), 或联邦), 效率可提高10 - 100倍, 取决于他们使用的编解码器.” Ronca acknowledged that “ASICs can definitely bend the curve down," but pointed out that “ASIC deployment of new codecs will lag SW by several years."

马克Donnigan, a marketing consultant and former Beamr VP of marketing, 假设内容自适应编码(CAE), or per-title encoding) techniques were a partial solution, pointing out that "there are some 'CAE' approaches that can reduce bitrate 30% on average, 只有20%的命中率可以计算, as opposed to a 1000% increase in the compute requirements when switching from a legacy standard to next-gen, VVC, 或以其他方式."

Other respondents wondered if the power savings from delivering lower-bitrate video offset the additional encoding cost. Alberto De Luca, distribution engineer at Discovery, Inc., 说, 然后从H264升级到H265, 对于相同的视频质量, 我们正在降低比特率. 所以传输的数据更少, 因此更少的电力消耗-总的来说, I’m led to think that one encoder using 10 Mbps in h265 is still 'greener' than the footprint of 25Mbps going across a WAN in H264."

As part of a long, complex reply, Ronca 说, "Within 5-10 years, 1Gbps bandwidth will be the norm. 在理论传输极限, it will take as much as 400M views of a video to justify the 100x compute going from x264 to vvenc." Resolving this particular issue goes far beyond my technical skills, 但这是一个有趣的论点. How many more cell towers or cable headends will we need to build (and power) if we
以200Mbps而不是40Mbps的速度传输8K视频?

Looking ahead, Jean-Baptiste Lorent, director of marketing and sales at IntoPIX, 推荐, "[C]omplexity requirements (in hardware and software) should be part of the call for requirements and acceptance criteria when you create a new standard." True, of course, but wasn’t that the goal of the Low Complexity Enhancement Video Codec (LCEVC)? 一个被认可了好几个月的标准, LCEVC显然已经实现了其效率目标, but has seen limited uptake among companies like Facebook, 在哪些地方节能才是真正重要的.

这对你来说意味着什么? Like recycling bottles and newspapers, the issue raised by Ronca should be considered by all of us. 在规划编码策略时考虑绿色. 优先考虑每个标题而不是实现新的编解码器. Consider ASIC-based hardware encoding over software encoding. 作为一个行业, let’s get a definitive answer as to the breakeven between power consumed by next-gen codecs and the power saved by lower-bandwidth delivery.