Where The Rubber Meets The Road: 2010 Encoder Comparison

However, Flix Pro’s ability to open multiple instances enabled it to harvest those extra cores, making it the multifile encoding champion, while Squeeze 5’s ability to open multiple instances helped close the gap with Episode in the eight file-encoding test. Adobe Media Encoder is acceptable for single-file encodes, but its inability to encode in parallel or open multiple instances makes it the overall slowest by a wide margin in multiple-file encoding tests.

I tested on the Windows platform with an HP Z800 liquid-cooled workstation with two 3.33GHz Quad-Core Nehalem Xeon processors and 24GB of RAM running 64-bit Windows Vista. Again, none of the encoding tools encoded in parallel, though as you can see, all encoders save Episode ran much faster on Windows than on the Mac, particularly benefiting Sorenson Squeeze in the eight-file encoding trials.

To test the enterprise level products, I encoded 16 1-minute files to VP6 format, using the HP Z800 for Rhozet Carbon Coder and the Mac Pro for Episode Pro. Note that I used the single-pass implementation for Carbon Coder, so the quality was slightly less than that offered by Episode Engine.

To summarize the VP6 encoding options, if you’re producing on the Mac, the Adobe Media Encoder delivers good quality results, but third-party tools lap it easily in terms of performance. Flix Pro is the best choice if VP6 is the only format that you deliver, while Episode delivers nearly the same single-file encoding speed and equivalent quality. If you’re encoding lots of VP6 and other files, consider Squeeze.

Adobe Media Encoder is much faster on Windows than on the Mac, but it still suffers competitively with multiple files. While Flix Pro is still the fastest alternative, Squeeze is slightly faster than Episode in single-file encoding, and it’s much, much faster with multiple-file encoding. In the enterprise encoding market, Episode Engine works best right out of the box; it’s faster and offers better quality than Carbon Coder. If you install Flix Exporter to boost Carbon Coder’s quality, expect encoding times to slow down by a factor of at least 20%–30%.

Producing H.264 Files
Where all VP6 encoding tools have their roots in code from On2, H.264 encoding tools utilize codecs from four different vendors, including Apple, Dicas, MainConcept, and Microsoft. In addition, where producing VP6 files is relatively straightforward (with few knobs and dials to twiddle), all H.264 encoding tools enable the control of a vastly different array of encoding parameters. These variables in codec origin and encoding controls makes for big differences in H.264 quality, which is why our H.264 encoder analysis will begin with a comparison of the quality of the various encoding tools.

Briefly, when comparing the output quality of these tools, I use two configurations, 640x480@500Kbps (468/32 audio) and 720p@928Kbps (800/128 audio), both at 29.97 and encoded using two-pass (or multipass) variable bitrate encoding. I encode using the High Profile, CABAC entropy encoding, a B-frame interval of three, with five reference frames and all search and similar encoding parameters set for maximum quality. When these options aren’t available, as with Apple’s Compressor, I use the highest quality settings enabled by the encoding tool. To assess quality, I compare still-frame quality, look for dropped frames that mar playback smoothness, and play each file in real time to see the extent of motion artifacts such as mosquitoes and background shifting.

In terms of quality, codec vendors MainConcept (Adobe Media Encoder, Sorenson Squeeze, Rhozet Carbon Coder) and Dicas (Telestream Episode) have been doing an interesting dance for the past couple of years. Specifically, first, MainConcept would pull ahead, then Dicas would catch up, then MainConcept would pull ahead again. In June 2009, the codecs were about even; by the end of the year, MainConcept was ahead again.

That said, while it’s noticeable in side-by-side comparisons, the difference isn’t commercially relevant because viewers rarely see side-by-side comparisons. On the other hand, Apple’s codec is a clear step behind and shouldn’t be used unless you have bandwidth to burn and don’t need the optimum data rate/quality mix. The big surprise this year was the quality of the H.264 encoding offered by Microsoft Expression Encoder, which is almost indistinguishable from MainConcept; it’s a pretty incredible performance for a rookie codec.

And, yes, for the record, let me state that the encoding quality produced by all the encoding tools that use the MainConcept codec is fairly uniform. Rhozet Carbon Coder certainly offers much greater access to the MainConcept encoding parameters, which is very useful for advanced compressionists needing to fine-tune their H.264 encoding for specific source footage. For most general-purpose footage, however, Adobe Media Encoder will produce quality very similar to that put out by Carbon Coder.

With H.264 quality under our belts, lets turn to encoding speed, starting on the Mac. Since Sorenson Squeeze worked fine for both H.264 and WMV, I’m including these times for comparison purposes.