The Pros and Cons of Software Decoding for Publishers and Viewers
A recent paper from Friedrich-Alexander University benchmarks energy consumption and compression efficiency for six video codecs across software and hardware decoders. While the study uses VP9 as a reference, the more relevant comparisons for today’s streaming environment are between H.264 and HEVC in hardware, and AV1 and VVC in software, reflecting real-world deployments.
The findings highlight trade-offs in compression efficiency and decoding energy usage that directly impact cost, battery life, and the viewer's experience. At the core is a question that hasn’t been clearly answered: Who is software decoding really for?
At the end of this article, we'll invite you to take a survey on how you--whether as a publisher, viewer, or both--are using and/or experiencing software decoding.
What the Data Shows
The paper from Friedrich-Alexander University evaluates energy consumption and compression efficiency for six codecs—H.264, HEVC, VP9, AV1, VVC, and AVM—using both software and hardware decoders. The testing included reference and optimized software decoders (with and without SIMD) and hardware decoding on an ARM-based single-board computer (SBC) using FFmpeg-accessible hardware blocks.
All codecs were tested using 1080p and 4K sequences, with both random access and low-delay configurations. Energy measurements were collected using internal CPU power meters (for software decoding) and external power instrumentation (for hardware decoding). Compression efficiency was calculated using Bjøntegaard Delta Rate (BDR), with VP9 as the anchor.
While the paper reports comparisons relative to VP9, we’ve focused on the combinations most relevant to actual deployment:
- · H.264 (hardware decode): Used universally for compatibility
- · HEVC (hardware decode): Common in premium and 4K workflows
- · AV1 (software decode): UGC deployment path on non-hardware-enabled devices
- · VVC (software decode): Still experimental, no hardware decode support yet
Table 1. Efficiency, power consumption, and battery impact of four tested configurations.
Here’s how to consider the data in Table 1, which shows the efficiency, power consumption, and battery impact of the four tested configurations. The baseline for most services is H.264, which plays in hardware. To reduce bandwidth costs, producers can also encode using HEVC, which enjoys near-ubiquitous hardware playback, or AV1 or VVC, which mostly is played in software.
As the table shows, AV1 provides around 33 percent better compression than HEVC but requires approximately twice the decoding energy when run in software. This is estimated to reduce playback time by 20 to 30 percent compared to HEVC in hardware.
VVC improves compression by about 14 percent over AV1 (and 42 percent over HEVC), but its software decoding energy use is nearly three times higher than HEVC decoded in hardware. This results in 35–50% shorter battery life.
For the record, in Table 1, compression efficiency is based on Bjøntegaard Delta Rate (BDR) results from the paper's Random-Access tests, normalized to H.264 at 100. Decoder energy use is based on BDDE results from the optimiized software decoders (SIMD on for AV1/VVC) and hardware decoders for H.264 and HEVC. Battery-life impact estimates assume decoder powers accounts for 30-40 percent of total device power draw during playback, scaled relative to H.264 hardware decoding.
For Publishers: Flexibility, Efficiency, and Cost Savings
Now let's explore how this data shapes the value proposition for publishers opting for software decoding and viewers that wittingly, or unwittingly watch services that deploy it.
For publishers, the key motivation for deploying AV1 (and VVC) is economic. Reduced bitrates translate to lower delivery costs, particularly at scale. But these codecs can also have a significant impact on viewer quality of experience, as a 1080p 3 Mbps stream encoded using AV1 looks better than a 720p 3 Mbps H.264 stream. Two recent examples show how publishers are leveraging these benefits.
Meta began delivering AV1 for Facebook Reels on iPhones in early 2022 (Figure 1). Within a week, average playback FB-MOS improved by 0.6 points, and average bitrates dropped by 12 percent. AV1 quickly accounted for around 70% of Reels' watch time on iPhones. Meta has since expanded its support to include HDR and 60 fps delivery on newer iPhones and select mid- and high-end Android devices.
Figure 1. The article at go2sm.com/reels details the path Meta took to deliver AV1 for software decode and the benefits it's delivering.
Netflix adopted AV1 to improve compression efficiency and expand reach to viewers on congested or capped networks. A/B tests demonstrated up to a 10-point VMAF improvement in challenging conditions, a 38% reduction in quality down-switching, and a 5% increase in 4K streaming hours. AV1 is now Netflix’s second most-streamed format, contributing to lower delivery costs and improved edge caching efficiency.
While the compression and delivery benefits of AV1 and VVC are real, they come with tradeoffs. Some are specific to software decoding; others apply more broadly to any next-generation codec.
The first is hardware-secure DRM. Platforms that require hardware-based content protection, such as Widevine L1 or FairPlay Secure, cannot use software decoding paths for encrypted or premium content because they do not meet the necessary security requirements. For publishers, the primary constraint with software decoding is the inability to support.
Playback consistency is another concern. Device capabilities vary, and software decode paths are more likely to suffer from stutter or instability. This variability is an issue Meta is actively working to address through VCAT (Figure 2), an Android/AV1 open-source benchmarking tool scheduled for Summer 2025 release that will benchmark AV1 playback for smoothness and battery impact (see go2sm.com/vcat). This will enable publishers to perform the benchmarking that Meta conducted before sending AV1 streams to Android devices for software decoding, as described below.
Figure 2. VCAT will help publishers benchmark the playback smoothness and battery impact of software AV1 playback (and later VVC) on Android devices.
Beyond decoding, all new codecs introduce operational complexity. Encoding workflows are slower and more resource-intensive, storage requirements can increase due to multi-encode strategies and testing and QA workloads grow with every added decode path. Caching efficiency may also suffer when content is fragmented across formats and resolutions.
Battery performance is another consideration, but it primarily affects the viewer. As discussed earlier, software decoding of AV1 and VVC on mobile devices can lead to shorter playback times compared to H.264 or HEVC hardware decode.
Choose Your Targets Wisely
Note that while Meta reported significant benefits from its AV1 deployment, it also invested heavily to prevent poor-quality playback experiences. By way of background, there are broad swaths of devices that can handle AV1 decode, including most iPhones in software, more recent iPhones in hardware, and their Android high-end equivalents. But particularly among low-cost Android phones, there's a great diversity in software-playback capabilities. In these devices, battery life isn't the issue; it's the potential for dropped frames, frequent stoppages, and an overall poor playback experience.
As Meta’s Reels article describes, while integrating AV1 on iOS was relatively straightforward, deployment on Android proved more complex. Meta had to test across a wide range of devices, implement logging to diagnose decoder issues across the playback stack, and develop custom performance scoring benchmarks to identify phones capable of decoding AV1 in real time. They also tuned their encoder presets to improve quality at lower resolutions, prefetched content to reduce startup delays, and implemented mixed-codec playback for compatibility with in-stream ABR switching.
This need for device-level gating was highlighted again in 2024, when YouTube appeared to shift some Android devices to software AV1 decoding. While the actual performance and battery impact may have been minor or isolated, the response was not. Users and reviewers flagged noticeable battery drain and playback issues, prompting widespread discussion.
Whether it was a molehill or a mountain, the takeaway is clear: without clear targeting and safeguards, even a technically sound deployment can damage perception and trust.
These challenges highlight the practical costs of deploying software decoding at scale, even when the codec offers clear efficiency and quality benefits. Integration complexity, performance variability, and the lack of standardized device profiling remain significant hurdles, particularly on platforms like Android with broad device fragmentation and many low-performance, sub-$100 devices. These challenges don’t preclude adoption, but they do shape where and how new codecs are deployed.
For Viewers: Battery Life, Quality, and Playback Stability
Most viewers do not know or care how their videos are decoded. But they care about the results, including battery life, video quality, and playback stability.
The Friedrich-Alexander University paper provides energy consumption data for software and hardware decoding. We used this data to estimate real-world battery life on a standard 4000 mAh mobile device.
Table 2. Power draw and estimated battery life of the listed decoding alternatives
Playback using software decoding can reduce battery life by 25 to 37 percent compared to hardware decoding. But battery life is not the only concern. Software decoding also increases the risk of frame drops, buffering, and inconsistent playback. These issues are especially common on lower-end or thermally constrained devices, where the CPU or GPU may not have enough available capacity to decode high-resolution or high-frame-rate video smoothly.
Thermal throttling is another concern. Software decoding generates more heat than hardware decoding, which can lead to reduced performance over time or an uncomfortably warm device. Viewers may not associate this with video playback directly, but the experience suffers just the same.
Multitasking also affects performance. Hardware decoding runs on dedicated silicon and is largely unaffected by other apps. Software decoding competes with every other process on the device, so performance can vary based on background activity.
From the viewer’s perspective, these problems may appear as random stutters or poor video quality. They may not know the root cause, but they will blame the app or the platform. This is why many streaming services continue to rely on hardware decoding wherever possible.
Hardware decoding offers more predictable results, better battery life, and fewer surprises across the device landscape.
User Benefits from Software Decode
While software decoding presents real challenges, it also enables access to more efficient codecs like AV1 and VVC, which can benefit viewers in multiple ways, as discussed in the Meta and Netflix mentions above. To recap, improved compression efficiency allows platforms to deliver higher-resolution video at lower bitrates, which helps reduce stalls and improve quality on limited connections. This is particularly important for users on mobile networks or in regions with bandwidth constraints.
For UGC platforms, improved encoding efficiency means more videos can be stored and streamed at higher quality without increasing infrastructure costs. That translates to more content, faster playback, and potentially better visual quality across the board. So, even though a viewer might not knowingly want software decoding, they likely want the benefits these codecs enable: more content, better quality, and smoother playback in bandwidth-constrained environments.
Take the Survey
Software decoding is not a universally better or worse option. It’s a tradeoff. For publishers, it opens the door to bandwidth savings and quality improvements, particularly with next-generation codecs such as AV1 and VVC. But it also introduces operational complexity, and limits use in DRM-protected workflows. For viewers, software decoding can reduce battery life and affect playback stability, but it also enables better quality on limited networks and unlocks more content, particularly from user-generated content (UGC) platforms.
Whether software decoding is the right approach depends on what you’re delivering, who you’re delivering it to, and how much control you have over the playback environment. We’ve laid out the data and real-world deployment examples. Now we want to hear from you.
We’re running two short surveys, and you’re welcome to take one or both:
- The publisher survey asks for your perspective on software decoding in your platform or service, including key drivers and concerns. If you're responsible for delivering video to users, your input will help us understand how the industry is approaching these tradeoffs.
- The viewer survey focuses on the experience side. If you watch a significant amount of video on your phone or tablet, we’d like to know what matters most to you in terms of quality, battery life, and playback smoothness.
If we reach critical mass for the surveys, we’ll share the results in a follow-up article. Thanks in advance for your time.
Related Articles
If you'll be encoding with SVT-AV1 or VVC, in this article you'll learn a bit about how to optimize your encodes, particularly the trade-offs that presets deliver, and how many logical processors to use. With SVT-AV1, I also explore the quality and playback efficiency of the fast decode option, although I didn't find much to show for it. You can also peruse the FFmpeg command strings used for all encodes.
08 May 2025
HEVC rode the 4K/HDR wave to success, but AV1, VVC, and LCEVC lack an equivalent killer app. With CDNs racing to the bottom on pricing and new royalty pools threatening to increase costs, codec adoption is no longer just about technical merit—it's about survival. These are the issues I'll explore in this article.
28 Mar 2025
The arc of AV1 codec adoption at large-scale content owners has been long and complex, as Meta Senior Media Software Engineer David Ronca (who also spent 12 years developing encoding solutions at Netflix) knows as well as anyone. In this interview with Jan Ozer of Streaming Media and Streaming Learning Center, Ronca discusses the integration of AV1 into the Android ecosystem, focusing on the challenges and solutions for mobile devices.
21 Feb 2025
What are the current innovations in reducing the carbon emissions and the carbon footprint of streaming? Jan Ozer of the Streaming Learning Center asks Kevin Yao of Amazon Web Services and David Ronca of Facebook about the innovations and new technologies their organizations are using to lessen the environmental impact of streaming.
30 Sep 2022