A Brief Analysis of the Three Mainstream HD Video Codec Technologies in the Market

As we all know, compared with standard-definition videos such as RMVB and DVD, the playback of high-definition videos undoubtedly requires higher computer hardware requirements.

1. Let's first look at how high-definition video improves requirements:

(I) The first is the increase in video traffic. This is the most intuitive improvement factor. We can refer to the table below. The table simply presents the traffic comparison of "video source", but in fact, the "different formats" of high-definition video determine the difference in traffic.

It can be seen that the data flow of DVD video is only about 9.5Mb/s, but the highest Blu-ray can reach more than 40Mb/s, which is more than 4 times higher. This is undoubtedly a major reason for the improvement in hardware configuration, because the amount of data that needs to be processed has increased a lot.

(ii) The second is the complexity of the encoding format. We know that each encoding format has its own algorithm. An excellent algorithm can compress the video to a smaller size, but restoring this algorithm requires stronger computing power. The H.264 high-definition encoding format is a typical representative. Perhaps some readers will encounter some high-definition videos, but the playback requirements are not high. It may be that it uses some simple algorithms, such as the Mpeg2 high-definition encoding format, but when encountering an encoding format similar to H.264, it will not be able to play smoothly.

Now it seems that H.264 has gained wide support due to its excellent algorithm. Although its algorithm is complex, it can achieve the highest compression while minimizing the loss of video details. Therefore, if you want to play high-definition videos smoothly in the future, H.264 encoding video is a test that must be passed. Of course, we cannot ignore VC-1 encoding, which is a coding format strongly promoted by Microsoft and has also been widely supported. Although its requirements are slightly lower, it also causes "playback difficulties" on many mainstream computers. As for Mpeg2 high-definition encoding, its requirements are very low and its future development prospects are not outstanding.

Let's take a look at how these codes are processed on computers. The processing of video encoding is generally divided into several steps, each step will complete the corresponding task, and these steps will also exist during decoding, so as to achieve smooth playback of the film. So who will handle these tasks? In the early days on computers, most of the steps of decoding and playback were handled by the CPU, that is, the processor software decoding (hereinafter referred to as soft decoding) playback, which is why the CPU usage rate is very high.

Simply put, any mainstream computer, even a few years ago, can play all the previous standard-definition videos without any problems using processor software decoding. Among high-definition videos, Mpeg2 high-definition encoded videos have the lowest requirements, followed by VC-1 encoded videos, and the most demanding H.264 high-definition videos may cause many old computers to be unable to play smoothly.

Although hardware is developing rapidly now, users can use high-end quad-core processors to control the CPU usage of HD soft decoding playback to less than 50%, but the price of such processors is very high, and the installation cost increases significantly. Although dual-core processors are much cheaper, the CPU usage of soft decoding playback may be very high, even to the point where it cannot be played smoothly.

If you want to reduce the burden on the CPU, you must have another accessory to take over the encoding processing work, and the graphics card is naturally the best choice, that is, the engine in the graphics card replaces the CPU to complete the video decoding processing task, thereby releasing the CPU load. As a general-purpose processor, the CPU has limited efficiency in video decoding, but the graphics card can directly integrate the video decoding engine into the hardware, thereby achieving extremely high efficiency.

[page]

2. HD video encoding process

So what are the steps of high-definition encoding, and which part has the highest load? Taking H.264 video as an example, it is divided into four main parts (see the figure below). The four blocks in the figure are basically the four most important steps of H.264 decoding, and also the four main parts of resource consumption. Among them, the first step "CAVLC/CABAC decoding" consumes the most computing resources, which is far higher than the other three steps (in short, CAVLC/CABAC are two different algorithms in the H.264 encoding specification, in order to improve the compression ratio, CABAC has a higher compression rate than CAVLC, but the decoding requirements are naturally higher).

3. HD video decoding process

Let's take a look at the decoding process comparison of the three mainstream encoding formats, including Mpeg2, VC-1, and H.264 (see the figure below). It can be seen that there are still many differences between the encoding formats, which is also the reason for the different requirements of the encoding formats. The H.264 encoding format is the most complex, so the system requirements are the highest, and VC-1 is slightly lower, but still much higher than Mpeg2.

Other encoding formats are similar to H.264.

[page]

4. Analysis of HD decoding resource consumption

So which step consumes the most CPU usage? The following test comparison should be the best explanation (see the figure below). The compared videos include Mpeg2 and H.264 (AVC), which involve all four steps we mentioned above. Obviously, "stream processing" is the most processor-intensive part of all encoding formats, but this part of Mpeg2 video does not cause much trouble because the CPU usage is less than 2%.

However, problems arise for H.264. The video stream processing at a 20Mb/s encoding rate will reach a considerable CPU usage rate. The 40Mb/s high encoding rate video is even more exaggerated. The CPU usage rate rises very quickly during the stream processing process. Combined with other processing, it is not surprising that the total CPU usage rate is very high.

As mentioned above, the processing components of these four steps are different, mainly the CPU and graphics card. In the past, the CPU had a high occupancy rate because it processed more steps. Therefore, if the graphics card can take on more decoding steps, the CPU can release more load to ensure smooth video playback.

The following figure shows the H.264 decoding process. If the graphics card does not take on any steps (the first line of the diagram), the CPU usage rate is very high, and the video cannot be played smoothly at all. If the graphics card can implement the decoding process of the last two steps (the second line of the diagram), the CPU can be partially freed up, but some graphics cards cannot implement the more critical and most loaded "stream processing", so the CPU usage rate has decreased, but it is still high.

With a graphics card that has an H.264 hardware decoding engine, all four processing steps of H.264 encoding can be completed (the third line in the diagram), that is, full decoding can be achieved. This is the fundamental reason why they can significantly reduce the CPU usage of H.264 HD video playback. Since the graphics card has completed all the HD decoding processing, the CPU is naturally idle. What we call partial decoding means that the CPU still undertakes certain processing tasks, so the usage rate is still much higher.