Easy & Safe Codec Modernization with Beamr using Nvidia GPUs
Following a decade where AVC/H.264 was the clear ruler of the video encoding world, the last years have seen many video coding options battling to conquer the video arena. For some insights on the race between modern coding standards you can check out our corresponding blog post.
Today we want to share how easy it can be to upgrade your content to a new and improved codec in a fast, fully automatic process which guarantees the visual quality of the content will not be harmed. This makes the switchover to newer encoders a smooth, easy and low cost process which can help accelerate the adoption of new standards such as HEVC and AV1. When this transformation is done using a combination of Beamr’s technology with the Nvidia NVENC encoder, using their recently released APIs, it becomes a particularly cutting-edge solution, enjoying the benefits of the leading solution in hardware AV1 encoding.
The benefit of switching to more modern codecs lies of course in the higher compression efficiency that they offer. While the extent of improvement is very dependent on the actual content, bitrates and encoders used, HEVC is considered to offer gains of 30%-50% over AVC, meaning that for the same quality you can spend up to 50% fewer bits. For AV1 this increase is generally a bit higher.. As more and more on-device support is added for these newer codecs, the advantage of utilizing them to reduce both storage and bandwidth is clear.
Generally speaking, performing such codec modernization involves some non-trivial steps.
First, you need to get access to the modern encoder you want to use, and know enough about it in order to configure the encoder correctly for your needs. Then you can proceed to encoding using one of the following approaches.
The first approach is to perform bit-rate driven encoding. One possibility is to use conservative bitrates, in which case the potential reduction in size will not be achieved. Another possibility is to set target bitrates that reflect the expected savings, in which case there is a risk of losing quality. For example, In an experimental test of files which were converted from their AVC source to HEVC, we found that on average, a bitrate reduction of 50% could be obtained when using the Beamr CABR codec modernization approach. However, when the same files were all brute-force encoded to HEVC at 50% reduced bitrate, using the same encoder and configuration, the quality took a hit for some of the files.
This example shows the full AVC source frame on top, with the transcodes to HEVC below it. Note the distortion in the blind HEVC encode, shown on the left, compared to the true-to-source video transformed with CABR on the right.
The second approach is to perform the transcode using a quality driven encode, for instance using the constant QP (Quantization Parameter) or CRF (Constant Rate Factor) encoding modes with conservative values, which will in all likelihood preserve the quality. However, in this case you are likely to unnecessarily “blow up” some of your files to much higher bitrates. For example, for the UGC content shown below, transcoding to HEVC using a software encoder and CRF set to 21 almost doubled the file size.
Yet another approach is to use a trial and error encode process for each file or even each scene, manually verifying that a good target encoding setup was selected which minimizes the bitrate while preserving the quality. This is of course an expensive and cumbersome process, and entirely unscalable.
By using Beamr CABR this is all done for you under the hood, in a fully automatic process, which makes optimized choices for each and every frame in your video, selecting the lowest bitrate that will still perfectly preserve the source visual quality. When performed using the Nvidia NVENC SDK with interfaces to Beamr’s CABR technology, this transformation is significantly accelerated and becomes even more cost effective.
The codec modernization flow is demonstrated for AVC to HEVC conversion in the above high-level block diagram. As shown here, the CABR controller interacts with NVENC, Nvidia’s hardware video encoder, using the new APIs Nvidia has created for this purpose. At the heart of the CABR controller lies Beamr’s Quality Measure, BQM, a unique, patented, Emmy award winning perceptual video quality measure. BQM has now been adapted and ported to the Nvidia GPU platform, resulting in significant acceleration of the optimization process .
The Beamr optimization technology can be used not only for codec modernization, but also to reduce bitrate of an input video, or of a target encode, while guaranteeing the perceptual quality is preserved, thus creating encodes with the same perceptual quality at lower bitrates or file sizes. In any and every usage of the Beamr CABR solution, size or bitrate are reduced as much as possible while each frame of the optimized encode is guaranteed to be perceptually identical to the reference. The codec modernization use case is particularly exciting as it puts the ability to migrate to more efficient and sophisticated codecs, previously used primarily by video experts, into the hands of any user with video content.
For more information please contact us at info@beamr.com
It has been two years since we published a comparison of the two leading HEVC software encoder SDKs; Beamr 5, and x265. In this article you will learn how Beamr’s AVC and HEVC software codec SDKs have widened the computing performance gap further over x264 and x265 for live broadcast quality streaming.
Why Performance Matters
With the performance of our AVC (Beamr 4) and HEVC (Beamr 5) software encoders improving several orders of magnitude over the 2017 results, it felt like the right time to refresh our benchmarks, this time with real-time operational data.
It’s no secret that x264 and x265 have benefited greatly, as open-source projects, from having thousands of developers working on the code. This is what makes x264 and x265 a very high bar to beat. Yet even with so many bright and talented engineers donating tens of thousands of development hours to the code base, the architectural constraints of how these encoders were built limit the performance on multicore processors as you will see in the data below.
Creative solutions have been developed which enable live encoding workflows to be built using open-source. But it’s no secret that they come with inherent flaws that include being overly compute intensive and encumbered with quality issues as a result of not being able to encode a full ABR stack, or even a single 4K profile, on a single machine.
The reason this matters is because the resolutions that video services deliver continues to increase. And as a result of exploding consumer demand for video, data consumption is consuming network bandwidth to the point that Cisco reports in 2022, 82% of Internet traffic will be video.
Cisco says in their Visual Networking Index that by 2022 SD resolution video will comprise just 23.4% of Internet video traffic, compared to the 60.2% of Internet video traffic that SD video consumed in 2017. What use to represent the middle-quality tier, 480p (SD), has now become the lowest rung of the ABR ladder for many video distributors.
1080p (HD) will makeup 56.2% of Internet video traffic by 2022, an increase from 36.1% in 2017. And if you thought the resolution expansion was going to end with HD, Cisco is claiming in 2022, 4K (UHD) will comprise 20.3% of all Internet-delivered video.
Live video services are projected to expand 15x between 2017 and 2022, meaning within the next three years, 17.1% of all Internet video traffic will be comprised of live streams.
These trends demonstrate the industry’s need to prepare for this shift to higher resolution video and real-time delivery with software encoding solutions that can meet the requirement for live broadcast quality 4K.
Blazing Software Performance on Multicore Processors
The Beamr 5 software encoder utilizes an advanced thread management architecture. This represents a key aspect of how we can achieve such fantastic speed over x265 at the same quality level.
x265 works by creating software threads and adding them to a thread pool where each task must wait its turn. In contrast, Beamr 5 tracks all the serial dependencies involved with the video coding tasks it must perform, and it creates small micro-tasks which are efficiently distributed across all of the CPU cores in the system. This allows the Beamr codec to utilize each available core at almost 100% capacity.
All tasks added to the Beamr codec thread pool may be executed immediately so that no hardware thread is wasted on tasks where the data is not yet available. Interestingly, under certain conditions, x265 can appear to have higher CPU utilization. But, this utilization includes software threads which are not doing any useful work. This means they are “active” but not processing data that is required for the encoding process.
Adding to Beamr encoders thread efficiency, we have implemented patented algorithms for more effective and efficient video encoding, including a fast motion estimation process and a heuristic early-termination algorithm which enables the encoder to reach a targeted quality using fewer compute resources (cycles). Furthermore, Beamr encoders utilize the latest AVX-512 SIMD instruction set for squeezing even more performance out of advanced CPUs.
The end result of the numerous optimizations found in the Beamr 4 (AVC) and Beamr 5 (HEVC) software encoders is that they are able to operate nearly twice as fast as x264 and x265 at the same quality, and with similar settings and tools utilization.
Video streaming services can benefit from this performance advantage in many ways, such as higher density (more channels per server) which reduces operational costs. To illustrate what this performance advantage can do for you- consider that at the top-end, Beamr 5 is able to encode 4K, 10-bit video at 60 FPS in real-time using just 9 Intel Xeon Scalable cores where x265 is unable to achieve this level of performance with any number of computing cores (at least on a single machine). And, as a result of being twice as efficient, Beamr 4 and Beamr 5 can deliver higher quality at the same computing envelope as x264 and x265.
The Test Results
For our test to be as real-world as possible, we devised two methodologies. In the first, we measured the compute performance of an HEVC ABR stack operating both Beamr 5 and x265 at live speed. And for the second test, our team measured the number of simultaneous live streams at 1080p, comparing Beamr 4 with x264, and Beamr 5 with x265; and for 4K comparing Beamr 5 with x265. All tests were run on a single machine.
Live HEVC ABR Stack: Number of ABR Profiles (Channels)
This test was designed to find the maximum number of full ABR channels which can be encoded live by Beamr 5 and x265 on an AWS EC2 c5.24xlarge instance.
Each AVC channel comprises 4 layers of 8-bit 60 FPS video starting from 1080p, and the HEVC channel comprises either 4 layers of 10-bit 60 FPS video (starting from 1080p), or 5 layers of 10-bit 60 FPS video (starting from 4K).
Live HEVC ABR Stack Test – CONFIGURATION
Platform:
AWS EC2 c5.24xlarge instance
Intel Xeon Scaleable Cascade Lake @ 3.6 GHz
48 cores, 96 threads
Presets:
Beamr 5: INSANELY_FAST
x265: ultrafast
Content: Netflix 10-bit 4Kp60 sample clips (DinnerScene and PierSeaside)
Encoded Frame Rate (all layers): 60 FPS
Encoded Bit Depth (all layers): 10-bit
Encoded Resolutions and Bitrates:
4Kp60@18000 Kbps (only in 4K ABR stack)
1080p60@3750 Kbps
720p60@2500 Kbps
576p60@1250 Kbps
360p@625 Kbps
Live HEVC ABR Stack Test – RESULTS
NOTES:
(1) When encoding 2 full ABR stacks with Beamr 5, 25% of the CPU is unused and available for other tasks.
(2) x265 cannot encode even a single 4K ABR stack channel at 60 FPS. The maximum FPS for the 4K layer of a single 4K ABR stack channel using x265 is 35 FPS.
Live AVC & HEVC Single-Resolution: Number of Channels (1080p & 4K)
In this test, we are trying to discover the maximum number of single-resolution 4K and HD channels that can be encoded live by Beamr 4 and Beamr 5 as compared with x264 and x265, on a c5.24xlarge instance. As with the Live ABR Channels test, the quality between the two encoders as measured by PSNR, SSIM and VMAF was always found to be equal, and in some cases better with Beamr 4 and Beamr 5 (see the “Quality Results” section below).
Live AVC Beamr 4 vs. x264 Channels Test – CONFIGURATION
Platform:
AWS EC2 c5.24xlarge instance
Intel Xeon Scaleable Cascade Lake @ 3.6 GHz
48 cores, 96 threads
Speeds / Presets:
Beamr 4: speed 3
x264: preset medium
Content: Netflix 10-bit 4Kp60 sample clips (DinnerScene and PierSeaside)
Encoded Frame Rate (all channels): 60 FPS
Encoded Bit Depth (all channels): 8-bit
Channel Resolutions and Bitrates:
1080p60@5000 Kbps
Live AVC Beamr 4 vs. x264 Channels Test – RESULTS
Live HEVC Beamr 5 vs. x265 Channels Test – CONFIGURATION
Platform:
AWS EC2 c5.24xlarge instance
Intel Xeon Scaleable Cascade Lake @ 3.6 GHz
48 cores, 96 threads
Speeds / Presets:
Beamr 5: INSANELY_FAST
x265: ultrafast
Content: Netflix 10-bit 4Kp60 sample clips (DinnerScene and PierSeaside)
Encoded Frame Rate (all channels): 60 FPS
Encoded Bit Depth (all channels): 10-bit
Channel Resolutions and Bitrates:
4K@18000 Kbps
1080p60@3750 Kbps
Live HEVC Beamr 5 vs. x265 Channels Test – RESULTS
NOTES:
(1) x265 was unable to reach 60 FPS for a single 4K channel, achieving just 35 FPS at comparable quality.
Quality Comparisons (PSNR, SSIM, VMAF)
Beamr 5 vs. x265
NOTES:
As previously referenced, x265 was unable to reach 4Kp60 and thus PSNR, SSIM, and VMAF scores could not be calculated, hence the ‘N/A’ designation in the 3840×2160 cells.
Video engineers are universally focused on the video encoding pillars of computing efficiency (performance), bitrate efficiency, and quality. Even as technology has enabled each of these pillars to advance with new tool sets, it’s well known that there is still a tradeoff between each that is required.
On one hand, bitrate efficiency requires tools that sap performance, and on the other hand, to reach a performance (speed) target, tools which could positively affect quality cannot be used without harming the performance characteristics of the encoding pipeline. As a result, many video encoding practitioners have adapted to the reality of these tradeoffs and simply accept them for what they are. Now, there is a solution…
The impact of adopting Beamr 4 for AVC and Beamr 5 for HEVC transcends a TCO calculation. With Beamr’s high-performance software encoders, services can achieve bitrate efficiency and performance, all without sacrificing video quality.
The use of Beamr 4 and Beamr 5 opens up improved UX with an increase in resolution or frame-rate which means it is now possible for everyone to stream higher quality video. As the competitive landscape for video delivery services continues to evolve, never has the need been greater for an AVC and HEVC codec implementation that can deliver the best of all three pillars: performance, bitrate efficiency, and quality. With the performance data presented above, it should be clear that Beamr 4 and Beamr 5 continue to be the codec implementations to beat.
Video engineers dedicated to engineering encoding technologies are highly skilled and hyper-focused on developing the foundation for future online media content. Such a limited pool of experts in this field creates a lot of opportunity for growth and development, it also means there must be a level of camaraderie and cooperation between different methodologies.
In past episodes, you’ve seen The Video Insiders compare codecs head-to-head and debate over their strengths and weaknesses. Today, they are tackling a deeper debate between encoding experts: the advantages and disadvantages of proprietary technology vs. community-driven open source.
In Episode 05, Tom Vaughan surprises The Video Insiders as he talks through his take on open source vs. proprietary technology.
Want to join the conversation? Reach out to TheVideoInsiders@beamr.com
TRANSCRIPTION (lightly edited to improve readability only)
Mark Donnigan: 00:00 In this episode, we talk with a video pioneer who drove a popular open source codec project before joining a commercial codec company. Trust me, you want to hear what he told us about proprietary technology, open source, IP licensing, and royalties.
Announcer: 00:18 The Video Insiders is the show that makes sense of all that is happening in the world of online video, as seen through the eyes of a second generation codec nerd and a marketing guy who knows what iframes and macroblocks are. Here are your hosts, Mark Donnigan and Dror Gill.
Mark Donnigan: 00:56 You know, we’re like encoding geeks. I mean, are there even 180 of us in the world?
Dror Gill: 01:01 I don’t know. I think you should count the number of people who come to Ben Wagoner’s compressionist breakfast at NAB, that’s about the whole industry, right?
Mark Donnigan: 01:09 Yeah. That’s the whole industry.
Mark Donnigan: 01:11 Hey, we want to thank, seriously in all seriousness, all the listeners who have been supporting us and we just really appreciate it. We have an amazing guest lined up for today. This is a little personal for me. It was IBC 2017, I had said something about a product that he was representing, driving, developing at the time. In fact, it was factually true. He didn’t like it so much and we exchanged some words. Here’s the ironic thing, this guy now works for us. Isn’t that amazing, Dror?
Mark Donnigan: 01:52 You know what, and we love each other. The story ended well, talk about a good Hollywood ending.
Mark Donnigan: 01:58 Well, we are talking today with Tom Vaughn. I’m going to let you introduce yourself. Tell the listeners about yourself.
Tom Vaughn: 02:10 Hey Mark, hey Dror. Good to be here.
Tom Vaughn: 02:12 As Mark mentioned, I’m Beamr’s VP of strategy. Joined Beamr in January this year. Before that I was Beamr’s, probably, primary competitor, the person who started and led the x265 project at MulticoreWare. We were fierce competitors, but we were always friendly and always friends. Got to know the Beamr team when Beamr first brought their image compression science from the photo industry to the video industry, which was three or four years ago. Really enjoyed collaborating with them and brainstorming and working with them, and we’ve always been allies in the fight to make new formats successful and deal with some of the structural issues in the industry.
Dror Gill: 03:02 Let me translate. New formats, that means HEVC. Structural issues, that means patent royalties.
Tom Vaughn: 03:13 Yeah, we had many discussions over the years about how to deal with the challenging macro environment in the codec space. I decided to join the winning team at Beamr this year, and it’s been fantastic.
Mark Donnigan: 03:28 Well, we’re so happy to have you aboard, Tom.
Mark Donnigan: 03:32 I’d like to just really jump in. You have a lot of expertise in the area of open source, and in the industry, there’s a lot of discussion and debate, and some would even say there’s religion, around open source versus proprietary technology, but you’ve been on both sides and I’d really like to jump into the conversation and have you give us a real quick primer as to what is open source.
Tom Vaughn: 04:01 Well, open source is kind of basic what it says is that you can get the full source code to that software. Now, there isn’t just one flavor of open source in terms of the software license that you get, there are many different open source licenses. Some have more restrictions and some have less restrictions on what you can do. There are some well known open source software programs and platforms, Linux is probably the most well known in the multimedia space, there’s FFmpeg and Libav. There’s VLC, the multimedia player. In the codec space, x264, x265, VP9, AV1, et cetera.
Dror Gill: 04:50 I think the main attraction of open source, I think, the main feature is that people from all over the world join together, collaborate, each one contributes their own piece, then somehow this is managed together. Every bug that is discovered, anyone can fix it, because the source is open. This creates kind of a community and together a piece of software is created that is much larger and more robust than anything that a single developer could do on his own.
Tom Vaughn: 05:23 Yeah, ideally the fact that the source code is open means that you have many sets of eyes, not only trying the program, but able to go through the source code and see exactly how it was written and therefore more code review can happen. On the collaboration side, you’re looking for volunteers, and if you can find and energize many, many people worldwide to become enthusiastic and devote time or get their companies motivated to allocate developers full- or part-time to a particular open source project, you get that collaboration from many different types of people with different individual use cases and motivations. There are patches submitted from many different people, but someone has to decide, does that patch get committed or are there problems with that? Should it be changed?
Tom Vaughn: 06:17 Designed by a committee isn’t always the optimal, so someone or some small group has to decide what should be included, what should be left out.
Dror Gill: 06:27 It’s interesting to see, actually, the difference between x264 and x265 in this respect, because x264, the open source implementation of x264 was led by a group of developers, really independent developers, and no single company was owning or leading the development of that open source project. However, with x265, which is the open source implementation of HEVC, your previous company, MulticoreWare, has taken the lead and devoted, I assume, most of the development resources that have gone into the open source development, most of the contributions came from that company, but it is still an open source project.
Tom Vaughn: 07:06 That’s right. x264 was started by some students at a French university, and when they were graduating, leaving the university, they convinced the university to enable them to take the code with them, essentially under an open source license. It was very much grassroots open source beginnings and execution where developers may come and go, but it was a community collaboration.
Tom Vaughn: 07:31 I started x265 at MulticoreWare with a couple of other individuals, and the way we started it was finding some commercial companies who expressed a strong interest in such a thing coming to life and who were early backers commercially. It was quite different. Then, because there’s a small team of full-time developers on it working 40 hours plus a week, that team is moving very fast, it’s organized, it’s within a company. There was less of a need for a community. While we did everything we could to attract more external contributors, attracting contributors is always a challenge of open source projects.
Mark Donnigan: 08:14 What I hear you saying, Tom, is it sounds like compared to the x264 project, the x265 project didn’t have as large of a independent group of contributors. Is that …?
Tom Vaughn: 08:29 Well, x264 was all independent contributors.
Tom Vaughn: 08:33 And still is, essentially. There are many companies that fund x264 developers explicitly. Chip companies will fund individual developers to optimize popular open source software projects for their instruction set. AVX, AVX2, AVX512, essentially, things like that.
Tom Vaughn: 08:58 HEVC is significantly more complex than AVC, and I think, if I recall correctly, x265 already has three times the number of commits than x264, even though it’s only been in existence for one third of the life.
Dror Gill: 09:12 So Tom, what’s interesting to me is everybody’s talking about open source software being almost synonymous with free software. Is open source really free? Is it the same?
Tom Vaughn: 09:23 It can be at times. One part depends on the license and the other part depends on how you’re using the software. For example, if it’s a very open license like Apache, or BSD, or UIUC, that’s an attribution only license, and you’re pretty much free to create modifications, incorporate the software in your own works and distribute the resulting system.
Tom Vaughn: 09:49 Software programs like x264 and x265 are licensed under the GNU GPL V2, that is an open source license that has a copyleft requirement. That means if you incorporate that in a larger work and distribute that larger work, you have to open source not only your modifications, but you have to open source the larger work. Most commercial companies don’t want to incorporate some open source software in their commercial product, and then have to open source the commercial product. The owners of the copyright of the GPL V2 code, x264 LLC or MulticoreWare, also offer a commercial license, meaning you get access to that software, not under the GNU GPL V2, but under a separate, different license, in which case for you, it’s not open source anymore. Your commercial license dictates what you can and can’t do. Generally that commercial license doesn’t include the copyleft requirement, so you can incorporate it in some commercial product and distribute that commercial product without open sourcing your commercial product.
Dror Gill: 10:54 Then you’re actually licensing that software as you would license it from a commercial company.
Tom Vaughn: 10:59 Exactly. In that case it’s not open source at all, it’s a commercial license.
Dror Gill: 11:04 It’s interesting what you said about the GPL, the fact that anything that you compile with it, create derivatives of, incorporate into your software, you need to open source those components that you integrate with as well. I think this is what triggered Steve Ballmer to say in 2001, he said something like, “Open source is a cancer that spreads throughout your company and eats your IP.” That was very interesting. I think he meant mostly GPL because of that requirement, but the interesting thing is that he said that in 2001, and in 2016 in an interview, he said, “I was wrong and I really love Linux.” Today Microsoft itself open sources a lot of its own development.
Mark Donnigan: 11:48 That’s right. Yeah, that’s right.
Mark Donnigan: 11:50 Well Tom, let’s … This has been an awesome discussion. Let’s bring it to a conclusion. When is proprietary technology the right choice and when is open source maybe the correct choice? Can you give the listeners some guidelines?
Tom Vaughn: 12:08 Sure, people are trying to solve problems. Engineers, companies are trying to build products and services, and they have to compete in their own business environment. Let’s say you’re a video service and you run a video business. The quality of that video and the efficiency that you can deliver that video matters a lot. We know what those advantages of open source are, and all things being equal, people gravitate towards open source a lot because engineers feel comfortable actually seeing the source code, being able to read through it, find bugs themselves if pushed to the limit.
Tom Vaughn: 12:45 At the end of the day, if an open source project can’t produce the winning implementation of something, you shouldn’t necessarily use it just because it’s open source. At the end of the day you have a business to run and what you want is the most performant libraries and platforms to build your business around. If you find that a proprietary implementation in the long run is more cost effective, more efficient, higher performance, and the company that is behind that proprietary implementation is solid and is going to be there for you and provide a contractual commitment to support you, there’s no reason to not choose some proprietary code to incorporate into your product or service.
Tom Vaughn: 13:32 When we’re talking about codecs, there are particular qualities I’m looking for, performance, how fast does it run? How efficiently does it utilize compute resources? How many cores do I need in my server to run this in real time? And compression efficiency, what kind of video quality can I get at a given bit rate under a given set of conditions? I don’t want the second best implementation, I want the best implementation of that standard, because at scale, I can save a lot of money if I have a more efficient implementation of that standard.
Mark Donnigan: 14:01 Those are excellent pointers. It just really comes back to we’re solving problems, right? It’s easy to get sucked into religious debates about some of these things, but at the end of the day we all have an obligation to do what’s right and what’s best for our companies, which includes selecting the best technology, what is going to do the best job at solving the problems.
Mark Donnigan: 14:24 Thank you again for joining us.
Mark Donnigan: 14:33 Well, we want to thank you the listener for, again, joining The Video Insiders. We hope you will subscribe. You can go to thevideoinsiders.com and you can stream from your browser, you can subscribe on iTunes. We’re on Spotify. We are on Google Play. We’re expanding every day.
Announcer: 14:57 Thank you for listening to The Video Insiders podcast, a production of Beamr Limited. To begin using Beamr’s codecs today, go to Beamr.com/free to receive up to 100 hours of no-cost HEVC and H.264 transcoding every month.
There should be no debate over which standard will be the next generation winner with Apple’s adoption of HEVC (H.265). As the leading HEVC commercial software encoder implementer, we see video distributors who were waiting on the sideline prior to Apple’s announcement jumping into evaluations now that Apple has released the public versions of iOS 11 and macOS High Sierra.
The criteria for determining the best HEVC implementation will vary across application and service type and the business model of the video service will influence the decision of which solution to use. Yet, there is one factor that should be at the top of the decision tree for software based video encoders, and that is the performance (speed) of the encoder.
If you operate the encoding function for a live streaming service, then the performance and speed of your encoder is already top of mind, and the benefits of a solution which is 2 times faster is understood. But for those who operate SVOD and VOD services, the need to measure performance may not be as obvious. Read on to discover the benefits of operating a software encoder that is 2x faster.
Why a speed “face off”?
There is an inherent tradeoff between speed and quality with video encoder’s. With unlimited computing resources, one simply needs to apply the appropriate CPU power to meet the quality objective being targeted. In the case that an encoder is slow and inefficient, provided it can deliver the desirable quality, one can always reach the quality target if sufficient computing power is available. But, in the real world, there are constraints that cannot be ignored- whether budget, power, or space in the rack, few services have the luxury of operating as if they have unlimited computing resources. Which is why CPU performance is as critical an evaluation vector as absolute quality.
With Beamr 5 you will experience the following operational benefits as a result of the encoding operation being up to two times faster than x265 (based on the settings applied).
Up to half the server resources compared with x265
50% of the power utility draw compared with x265
50% less cooling and connectivity cost
Half the rack space
Evaluating Software Video Encoders
First, let’s examine the correlation between speed and quality by taking a look at the world’s best HEVC software video encoder, Beamr 5, as compared to the free open source alternative, x265.
The complexity of video encoder evaluations is such that you must understand not only what video files were used for the input, but also the precise settings (configuration) used. For this speed and quality evaluation you will note that x265 offers ten (10) performance presets that do a reasonable job of balancing the encoder’s speed and efficiency. Beamr 5 has a similar parameter named ‘enc_speed’ where there are a total of sixteen (16) levels available.
Out of the box Beamr 5 offers more granularity in its speed selection than x265, allowing for easier and more precise tuning across a wider range of computing platforms and architectures. However, to ensure an “apples to apples” comparison, other than matching the speed preset of each encoder, all other default parameters were left unchanged.
We want you to experience Beamr 5’s speed and quality first hand which is why we’ve provided the x265 settings that were used for each test. Readers are encouraged to duplicate the same tests, and we have provided download links for all files so that you can run your own independent analysis. Don’t believe our results? You can check them yourself.
As a commercial vendor working with the largest OTT and broadcast customers in the world, we design our products to be maximally usable by any video encoding engineer operating a video service. For the analysis we used x265 ‘veryslow’, ‘medium’, and ‘ultrafast’ settings since this test was designed to mirror real world use cases and not theoretical targets of either extreme speed or quality.
At Beamr we are all about the numbers, but as all video encoding engineers say, the eyes never lie. After you review the data in the charts below, please download the corresponding files so you can see for yourself Beamr 5’s superior quality.
We start with comparing the highest video quality modes of both encoders, akin to what an SVOD service may use, and move to slightly reduced video quality, but with faster performance needed for real-time (live) applications. Here we can demonstrate that Beamr 5 ‘enc_speed 0’ is much faster than the ‘veryslow’ x265 preset, while at the same time Beamr 5 produces better overall video quality.
The original source files can be downloaded from https://media.xiph.org/video/derf/, and converted from .y4m to yuv using the following FFmpeg command:
All Beamr 5 and x265 encoded files may be download from the links in the following tables or by click here. (all files were encoded in 4K resolution)
To aid in your subjective visual evaluation we suggest that you use a video comparison tool that is capable of rendering two videos in sync on the same screen. If you do not have a copy of Beamr View (formerly VCT), you may request a copy here. Beamr provides a limited license to our PC and Mac HEVC and H.264 software viewing tool FREE of charge to qualified video distribution services, content owners, and video platforms.
Test machine spec used for all clips: Amazon EC2 c3.4xlarge instance, Intel Xeon E5-2680 v2 @ 2.8GHz (16 threads), 30GB RAM, 160GB SSD.
The following chart provides a comparison of CPU performance at the maximum encoding quality with x265 configured at the ‘veryslow’ setting and Beamr 5 working at speed 0. The following chart shows the ability of Beamr 5 to produce higher video quality with much greater speed than x265. All other configuration settings were unchanged, except those required to align both codecs in keyframe interval, rate control limitations, and appropriate multithreading settings.
Table 1: Highest quality (VOD) configuration.
These are the settings used to test the highest quality presets (Beamr 5 enc_speed 0, x265 preset ‘veryslow’)
Command line example for x265 Highest Quality VOD setting:
4K resolution video ‘Ritual Dance’ frame 166, left side is Beamr 5, right side is x265. Beamr 5 is 71% faster than x265 with noticeably better quality.
4K resolution video ‘Driving POV’ frame 234, left side is Beamr 5, right side is x265. Beamr 5 is almost twice as fast as x265 at 98% with demonstrably better video quality.
With x265 set to the ‘medium’ preset and Beamr 5 ‘enc_speed’ set to 3, we observed that Beamr 5 maintained a 15% to 50% speed advantage over x265 while Beamr 5 consistently produced better quality. As with all other tests, configuration settings were unchanged, except those required to align both codecs in keyframe interval, rate control limitations, and appropriate multithreading settings.
Table 3: High quality preset configuration.
Command line example for x265 High Quality setting:
Table 4: High quality preset performance Beamr 5 vs. x265.
Screen capture 4: High quality settings Beamr 5 vs. x265.
4K resolution video ‘Wind and Nature’ frame 518, left side is Beamr 5, right side is x265. Beamr 5 is 50% faster with a slight edge on video quality over x265.
Screen capture 5: High quality settings Beamr 5 vs. x265.
4K resolution video ‘Ritual Dance’ frame 565, left side is Beamr 5, right side is x265. Beamr 5 is 50% faster than x265 and Beamr 5 has a definite edge on video quality.
Screen capture 6: High quality settings Beamr 5 vs. x265.
4K resolution video ‘Pier Seaside’ frame 18, left side is Beamr 5, right side is x265. Beamr 5 is 43% faster than x265 and Beamr 5 has noticeably better video quality.
Let’s examine how x265 and Beamr 5 fair when used for live encoding of 4K resolution video. When we set Beamr 5 to an equal speed configuration of x265 ‘ultrafast’, Beamr 5 was found to be faster while producing the same or better video quality.
Table 5: Live preset configuration.
Command line example for x265 live (real-time) configuration:
Table 6: Live configuration performance of Beamr 5 vs. x265.
Screen capture 7: Live settings Beamr 5 vs. x265.
4K resolution video ‘Bar Scene’ frame 595, left side is Beamr 5, right side is x265. Beamr 5 is 15% faster and produced better video quality than x265.
4K resolution video ‘Dinner Scene’ frame 528, left side is Beamr 5, right side is x265. Beamr 5 is 15% faster than x265, while both exhibited comparable quality.
Beamr has 30 granted and 23 pending patents, including a fast motion estimation process which is useful when speed and quality are at competing priority. Beamr 5’s fast motion estimation process is a significant factor in our performance advantage.
Beamr 5 uses a heuristic early-termination process which enables it to reach a targeted quality using less computational resources.
Beamr 5 allows for more possible frame hierarchies (levels) than what is provided by the H.265 standard. When encoding hierarchical B-frames, x265 uses a 2-level B-frame hierarchy. Beamr 5 on the other hand, encodes with a 3-level B-frame hierarchy when encoding seven B frames between P-frames.
x265 begins encoding frames before all reference frames are finished, which limits motion estimation to only the parts of reference frames that are available. When high performance (speed) is needed, x265 encoding quality can be degraded. This provides a serious advantage to Beamr 5 which supports full codec multithreading while x265 uses slices and tiles for parallelism (when WPP is disabled for maximum efficiency).
Beamr 5 achieves better parallelism without the limitations noted above because it is based on utilizing all possible independent encoding tasks inside a single frame. This is called wave-front and it is active even when entropy wave-front is not used. Multithreading in de-blocking and SAO filtering provides additional quality benefits.
Evaluating an encoder is one of the most important jobs for a video encoding engineer and we are proud of the advanced HEVC codec implementation we’ve developed. All information presented may be tested and verified and we invite you to take advantage of the video files using Beamr View to subjectively analyze the results above.
![In the battle between open source & proprietary technology, does video win? [podcast]](https://blog.beamr.com/wp-content/uploads/2019/01/Beamr-Podcast-banner-E2-1-1-800x544.jpg)
