High-End Video Transcoder Shootout: Elemental Vs. Telestream
We've upgraded our testing methodology to accurately reflect what broadcasters and other high-volume video publishers ask of their encoding systems. Our first subjects: Elemental Server and Telestream Vantage Lightspeed Server.
Learn more about the companies mentioned in this article in the Sourcebook:
High-end transcoding tools cost anywhere from $10,000 for a software-only system to “If you have to ask, you can’t afford it.” They’re the lifeblood of organizations that encode high volumes of daily, weekly, or monthly videos. In this article, we describe the factors you should consider when purchasing such a high-end system and present quality and performance comparisons of two systems, Elemental Technologies’ Elemental Server, and Telestream’s Vantage Lightspeed Server.
Why only two? Because, unfortunately, the encoding world doesn’t revolve around our editorial calendar. This process started with four companies, but two dropped out because they were close to releasing product updates that would make the results obsolete by the time they appeared in print. As you’ll learn, this review cycle incorporates a new performance and test routine which we will use to perform follow up reviews on additional systems as they come to market.
What factors should you consider when purchasing a high-end system? In 2013, Telestream sponsored a transcoder survey of the Streaming Media audience and was kind enough to share some of the results with us. The survey had 987 total respondents, 18 percent of whom spent from $10,000 to $50,000, with 16 percent spending $50,000 or more. This review is directed at that group.
What did these respondents care about? This is shown in Figure 1. Not surprisingly, at 26.4 percent, format support was the top consideration; if the system can’t deliver the required formats, it’s immediately out of the running. Next up is image quality, at 26.2 percent, and we performed extensive subjective and objective quality comparisons for this article.
Figure 1. Top considerations when making a transcoding purchase decision
Reliability is next at 17.64 percent, though this is tough to test in a relatively short test cycle. That said, both systems use industry standard, off-the-shelf hardware from vendors such as Intel and NVIDIA. That hardware is unlikely to fail, and both systems run very mature and market-proven preprocessing and encoding routines. We ran both through more than 3 days of continual testing that likely would have revealed any software-related flaws, and there were none.
Price and value was next at 19.47 percent; since only one vendor shared pricing information, you’ll have to gather your own input for this one.
Performance sits next at 5.91 percent; though speed is easily measured, and therefore prominent in many product comparisons, small variations don’t appear to move the buying needle. Both systems in this review were excellent performers, and the results were relatively close. Presumably if a system were orders of magnitude slower, it could dramatically impact the buying decision, but that wasn’t the case here.
I’ll add two more factors: First, you should consider the overall product offering. Specifically, where Elemental Server is a stand-alone transcoding engine, Vantage Lightspeed is a component of the Vantage workflow system, which incorporates a range of additional features, from integrated pre- and post-encoding quality control (QC) (which costs extra) to branched encoding logic that can integrate file headers and direct incoming files into specific encoding buckets. While you can certainly incorporate third-party QC with Elemental Server, if you’re seeking a workflow system, Telestream has the advantage.
You should also consider the integration and maturity of each company’s cloud encoder, since many companies purchasing an encoder in this class will need cloud encoding during its useful life. Elemental Cloud is mature and very well integrated, and I’ve tested several product generations with very good results. Telestream has a cloud offering that I’ve never tested, so you’ll need to make your own inquiries.
Our Test Systems
Let’s take a look at the two tested systems and then move to our updated quality and performance tests. As an overview, we tested the fastest system each vendor could provide in a single box. For Elemental, this was the Elemental Server, and our test system was configured with a 2.7 GHz 12-core (24 with HTT) Intel Xeon E5-2697 v2 CPU, with two NVIDIA K10 Tesla GPUs, with 16GB of RAM and running CentOS. Elemental Technologies declined to provide current pricing for the system, though the last time we looked, in May 2012, the system cost $25,950.
For Telestream, this was the Lightspeed Server, which ran two 2.5 GHz 10-core (20 with HTT) E5-2670 v2 Xeon CPUs, with two NVIDIA Tesla K20 Kepler-based GPU cards with 5GB RAM each, running Windows Server 2008 R2 Standard 64-bit on 32GB of system RAM. Lightspeed Server is priced based on the feature set and can cost as much as $36,000.
Our Updated Tests
To provide insight into our goals for these updated tests, I’ll briefly describe their genesis. Once we decided to run these tests, I asked the participants for comments about our previous testing. In a response that typified the frankness of the results, one participant replied, “Since we’re calling this a high-end encoder test, would you feel comfortable truly making it such? Right now the way the testing is set up, it really only applies to the less than 10 percent of customers who use high-end encoders in practice. It’s not testing products the way that our customers use them. It’s more of a videographer’s method of doing things.”
We took this and similar comments to heart. Without addressing all suggestions—for example, we didn’t incorporate captions or multilanguage audio into our tests—we did dramatically expand test duration and output formats. For output support, we produced 24 outputs from most test files, 12 MP4 files, as if destined for RTMP-based adaptive streaming, and 12 outputs for HTTP Live Streaming (HLS), as if for Apple, Android, and other compatible devices. We based the adaptive group on Apple’s TN2224, with expanded options at the high end to represent the typical grouping produced by customers of the participants. You can view the complete spec sheet at (go2sm.com/ozerencodespecs); a summary of the basic parameters is shown in Table 1.
Table 1. High-level configuration options for our adaptive group
We encoded all files using 1-pass 125 percent constrained VBR and defined most key parameters available in all systems, such as B-frames (three), reference frames (five), and audio parameters (96Kbps stereo, 44.1 kHz). Beyond these defined parameters, each vendor was free to configure its encoding parameters as it wished, subject to the proviso that the same preset had to be used for both quality and performance tests. For most tests, each encoder produced one set of 12 MP4 files and another set of 12 HLS streams.
In terms of operation, rather than have the vendors ship the systems to me, I ran both remotely, Elemental via the system’s web interface, and Vantage via Microsoft Remote Desktop. This afforded everyone easy access to the system, which was necessary to make the small changes to presets and other parameters essential to such a review.
To test performance, we ran these three tests:
- Television episode encoding. Here we tested 25 clips in MXF format averaging about 40 minutes in duration encoded to the 24 outputs.
- Advertising encoding. Here we encoded 200 30-second clips in MXF format to the adaptive group.
- 4K testing. Here we encoded 10 4K versions of the Blender Foundation movie Tears of Steel in H.264 format to the adaptive group.
As mentioned above, the vendors created all presets, which I reviewed. Before running any tests, I tested the presets with a 3-minute file to ensure that each vendor was hitting the target data rate within 5 percent. Table 2 shows the average time of two difference tests, plus the difference in performance between the two vendors.
Table 2. Performance tests and performance deltails
As the table shows, Elemental was about 25 percent faster in two of the three tests, with Telestream performing slightly faster in the commercial testing. To put the first test in perspective, at an average duration of 40 minutes, it involved converting the 25 files into about 400 hours of compressed video, which both systems accomplished comfortably in less than half a day.
Elemental is a strong choice for high-volume video publishers or those already using Elemental hardware, but it doesn't come at a bargain price.
The newest version of this popular live video production tool offers significant improvements and is a no-brainer upgrade for existing customers.
For broadcasters and high-volume producers, Amazon's Elastic Transcoder has too many limitations. For everyone else, it's an appealing, if flawed, solution.
When an encoding job is highly demanding, it takes powerful gear to create and judge all the files. Here are the right tools for the job.
At NAB, Elemental will show a bandwidth-saving solution that lets operators simplify creating multi-codec multi-delivery video.
With VOD Producer DAI, content owners can automatically create DAI-complaint files within their existing cable infrastructure.