HEVC vs. H.264: Bandwidth and Cost Savings
The case for moving from ubiquitous older video codecs like H.264 to newer codecs like AV1 or HEVC (H.265) is typically expressed in terms of encoding efficiency that translates to bandwidth and cost savings. For major content companies like Warner Bros. Discovery that have adopted H.265, how much has their experience borne out that hypothesis?
Page 1
The case for moving from ubiquitous older video codecs like H.264 (AVC) to newer codecs like AV1 or HEVC (H.265) is typically expressed in terms of encoding efficiency that translates to bandwidth and cost savings. For major content companies like Warner Bros. Discovery that have adopted H.265, how much has their experience borne out that hypothesis? In this clip from Streaming Media Connect 2023, Streaming Learning Center’s Jan Ozer and Alex Zambelli, Technical Product Manager at Warner Bros. Discovery, discuss the real-world ramifications of H.264-to-H.265 migration, where the realized payoff has proven significant, and where it’s been less noticeable. As it turns out, adopting a shiny new codec for significantly less-than-4K encoding and delivery isn’t always worth it when averaged out over all the rungs on the encoding ladder.
In response to Ozer’s open-ended question on where WBD is seeing savings with HEVC over H.264, Zambelli explains that it varies “depending on resolution. When you're looking at HD resolutions and 4K resolutions, it's easy sometimes to see savings between 25 and 40% with HEVC over AVC. And so some of those promises of varied savings that we expected from HEVC do come true at the higher resolutions. But then as you go down in resolution and you still have to support a full encoding ladder, and so often you find yourself encoding 320x180 video, the bitrate savings are not as huge. And so I would say across the board, the bitrate savings end up averaging out probably closer to 15 or 20% over AVC.
To get at the actual cost savings—which only partially increase in proportion to the bandwidth savings, since other factors come into play—Ozer asks Zambelli to consider and comment on two costs—“encoding both formats and the caching efficiency”—and how they factor in.
It turns out that the increased compute costs associated with HEVC encoding diminish to some extent the benefits of migrating to the newer codec, although the shift to cloud computing and improved encoders are helping to level the playing field. "There are certainly higher compute costs for HEVC," Zambelli says, "though I think the encoders have gotten a lot better over the last several years. And most cloud computing platforms have improved in terms of their CPU offerings, so we have highly multi-threaded CPUs available in the cloud. Most encoding these days is parallelized as well," he explains. "So if you're building a video transcoding platform that needs to handle a very high volume of content, you're going to parallelize it so that you're not necessarily encoding one video in one machine; you're going to spread it out across multiple machines. And so when we do it that way, the compute costs increase from HEVC. It's certainly there, but it's not so significant that it would prevent us from doing HEVC," he explains. "And so we're definitely reaching a point where I think HEVC encoding is going to become cheap enough that it'll almost seem like a free add-on, on top of the AVC encoding."
As for the fragmentation of caching, Zambelli says, “I haven't seen much of an impact from that, because all of our devices support AVC, and so obviously that's going to be highly cached. Over 75% of the devices support HEVC, and so we're getting a lot of caching efficiency for HEVC as well. So I would say we haven't seen any negative impact on quality of experience due to caching fragmentation.”
Getting right to the bottom line, Ozer asks, “What about cost? Aren’t you paying for two times the caching?”
Zambelli’s response: it’s complicated. “I feel like a lot of times it's assumed that there’s a direct one-to-one mapping between CDN cost savings and the rate savings from the more advanced codecs. And what I've found is that that's simply not the case,” he says. “Because if everybody in the world had unlimited bandwidth, if we all had unlimited connection speeds, then that would perhaps be true. But in reality, if you reduce the barrier, for example, of your 1080p encode from 6 megabits per second in AVC down to, let's say, 4.5 Mbps in HEVC, somebody who’s in the three-megabit connection is still going to get three-megabit video. So you’re not necessarily reducing the amount of data you’re delivering to those people. And as it turns out, there's still a lot of the population—not just in the US but certainly around the world—that has connection speeds that are below four or five megabits a second. And so real CDN savings end up being probably closer to 5 or 10% if you’re saving 20% on bitrate overall. So it’s definitely not a one-to-one mapping between those two.”
Learn more about the the state of video codecs at the next Streaming Media Connect in November 2023.
Page 1
Related Articles
In this article, I examine low-latency services provided by streaming CDNs, removing actual vendor names and making conclusions based on delivery method. I'll start with a discussion of the basics of latency before moving on to the testing results and analysis.