HDR 4K Player: Hardware Decoding vs Software Decoding

– A Case Study of 5KPlayer

Abstract: This report of a control experiment intends to examine high quality video playback performance in a software decoding environment and a hardware decoding environment with DearMob's hardware accelerated 5KPlayer as the subject. The result suggests a sharp reduction of CPU workload with the introduction of hardware acceleration shipped 5KPlayer to boost high resolution video playback.

Keywords: Hardware acceleration; software decoding; hardware accelerated video player; 5KPlayer.

Introduction

The debate over whether hardware acceleration or software decoding suits graphic processing tasks better has been on for a while. This page attempts to accurate the comparison to data with experimental figures.

Hardware acceleration technology refers to the use of computer hardware units to share the workload of CPU in such highly-demanding tasks as video decoding and rendering. Amongst the massive applications of hardware acceleration technology in the field of graphic processing, the decoding of high resolution video files relies on hardware accelerated video players such as 5KPlayer for more effective rendition of 4K videos, HDR videos, etc.

This article is an experiment-based report to illustrate 1080p, 4K, and 4K HDR video decoding and rendering performance under pure software decoding environment and hardware acceleration enabled environment - using 5KPlayer – the free video player for Windows and Mac as the subject.

Choosing 5KPlayer As the Experimental Subject

This experiment stays with 5KPlayer's hardware acceleration technology to switch between the software decoding mode and hardware accelerated decoding mode. For personal computers shipped with qualified graphic processing units (GPU), 3 optional modes are accessible to speed up the processing of graphics: Intel Quick Sync Video mode, NVIDIA CUDA mode, and DXVA 2.0 mode – to enable smooth decoding and rendering of H.264 videos.

1. Solid software kernel framework of 5KPlayer makes the experiment possible. This hardware acceleration video player for Windows and Mac sporting native compatibility for 1080p, 4K UHD, and 4K HDR videos enables relative data to be generated at first hand: When the results could only be arrived at through the comparison of CPU occupation under pure software decoding environment and under the hardware acceleration enabled environment, reaching a conclusion is out of the question should a software that crashes upon 4320p H.264 video files was adopted.

2. Relatively wide hardware acceleration mode compatibility – to decode and render high resolution videos – makes 5KPlayer an ideal subject as a free HD video player: modes available to be enabled include Intel Quick Sync Video Generation 1,2,3,4 (Sandy Bridge, Ivy Bridge, Haswell and Broadwell), NVIDIA CUDA (part of NVIDIA CUVID decoding technology), and Microsoft DXVA 2.0.

Preparations of the Experiment

Specifications of the Experimented Personal Computers:

PC-1: Windows 10 pro 64bit (DirectX 12)
CPU: Intel(R) Core(TM) i5-4460 CPU@3.20GHz
GPU: Intel HD graphics 4600
Internal HD: WDC WD10EZEX-00BN5A0 (1TB/7200 rpm)
RAM: 16G Kingston ddr3 \Samsung SAM0B26
Monitor: Samsung SAM0B26 S27D360 (27.2 inches)
PC-2: Windows 10 64bit (DirectX 12)
CPU: Intel Core i7-8700K @3.70GHz 6 cores, 12 threads
GPU: Nvidia GeForce GTX TITAN X (12GB/Nvidia)
RAM: 16G (Kingston DDR4 2400MHz)
Internal HD: WDC WD10EZEX-08WN4A0 (1TB/7200 RPM)
Monitor: Samsung SAM08E0 S27B370 (27 inches)

Specifications of the Experimented Video Samples:

Sample-1: 1080p 8bits video: MPEG-4 H264 AVC, 23.976 FPS, overall bit rate: 1564 kb/s
Sample-2: 4K 8bits video: MPEG-4 H264 AVC, 25.000 FPS, overall bit rate: 13.1 Mb/s
Sample-3: 4K HDR 8bits video: MPEG-4 H264 AVC, 59.940 FPS, overall bit rate: 94.8 Mb/s

Design of the 2 Parts of the Experiment:
This experiment with the free hardware accelerated video player for Windows 10 unfolds into 2 parts, with CPU workload data in software decoding mode, hardware accelerated decoding mode, and hardware accelerate rendering mode recorded in table -1 and table-2 given below.
Part-1: Playing video sample 1,2, and 3 on PC-1, first with software decoding only, and then with hardware acceleration enabled on 5KPlayer.
Part-2: Playing video sample 1,2, and 3 on PC-2, first with software decoding only, and then with hardware acceleration enabled on 5KPlayer.

Experiment Part-1: Software Decoding vs Hardware Decoding on PC-1

With the data collected below, it's not hard to see that the computer enjoys an obvious declination on CPU consumption by two-thirds when hardware acceleration is turned on, from 13% and 40% to 4% and 13% for HD 1080P video and 4K video playback respectively. Hardware acceleration also boils down GPU usage for video rendering to around 10%, the fact of which is more clearly seen on higher-res 4K video rendering - from 40% GPU usage to 13%. This is quite a bless for those multi-tasking lovers who are doing online shopping, photo editing, video chatting, playing small games, and desire to watch an HD/4K video at the same time.

On PC-1, the actual visual result for HDR 4K 8bits video playback is a total mess when hardware acceleration is disabled - video images are skipping frames, generating mosaics and freezing screen. Though the CPU usage data is relatively lower during software decoding, the video is not at all viewable. However, when hardware acceleration is enabled, 4K HDR 8bit video plays back smoothly instantly. And DXVA 2.0 decoding of 5KPlayer works notably better than QSV decoding in terms of smoothness, with less consumption of CPU and GPU regarding decoding and rendering alike.

table-1

Experiment Part-2: Software Decoding vs Hardware Decoding on PC-2

Drawn from the above statistics, hardware acceleration's effect on CPU/GPU consumption for 4K/HDR video playback is much noticeable than that for HD video playback. Hardware acceleration paired with an aced GPU can greatly relief CPU consumption and boost video decoding and GPU rendering efficiency to one-third, presenting a smoother visual enjoyment.

Again, the higher your video resolution is, the clearer hardware acceleration's efficiency is appreciated during video playback. And the adoption of hardware acceleration becomes more necessary following the resolution evolution. Hardware acceleration for HD/4K/HDR video playback aptly relives CPU usage and perfectly manifests visuals without glitches.

table-2

Tips and Notes:
The above 3 video samples: 1080p, 4K and 4K HDR videos are downloaded via 5KPlayer from online video channel. The 2 corresponding links are given here in case you'd like to go through a test yourself.
Address for 1080p Thunder Imagine Dragons: https://www.youtube.com/watch?v=fKopy74weus
Address for 4K HDR The World in HDR in 4K: https://www.youtube.com/watch?v=tO01J-M3g0U&t=6s.

Results: Hardware Accelerated Video Player Experiment

1: Comparison of 4K HDR Video (Sample-3) Decoding on PC-1 and PC-2

When playing the 4K HDR video sample, 5KPlayer can decode and smoothly render 4K HDR video in the high configuration computer PC-2, while lower-power PC-1 is not so competent enough to complete the video decoding and rendering of 4K HDR video alone when using software decoding only.

The experiment suggests that on PC-1, the actual visual result for HDR 4K 8bits video (sample -3 )playback is a total mess when hardware acceleration of 5KPlayer is disabled - video images are skipping frames, generating mosaics and freezing screen. Though the CPU usage data is relatively lower during software decoding, the video is not at all viewable.

However, when hardware acceleration is enabled, 4K HDR 8bit video plays back smoothly instantly. And DXVA 2.0 decoding of 5KPlayer works notably better than QSV decoding in terms of smoothness, with less consumption of CPU and GPU regarding decoding and rendering alike.

Despite that there is no specific data to prove how efficient hardware acceleration is to help 4K HDR videos smoothly on low-end computer PC-1, it is the truth that 4K HDR playback works smoothly on such PC when hardware acceleration is activated. Besides, the consumption of CPU and GPU saves a lot on highly-equipped computer PC-2 if GPU acceleration of 5KPlayer is enabled.

2: CPU Workload Comparison of 1080P, 4K, 4K HDR Video Samples via Software Decoding and Hardware Decoding

As a 4K HD video player, 5KPlayer can play 1080p HD and 4K videos naturally by default without turning on the hardware acceleration. But, the fact is that your computer will go much smoother when there are multiple tasks running on computer if the hardware decoding mode is enabled. 5KPlayer's hardware acceleration utility is able to allocate a major part of the heavy decoding task from the CPU to the GPU to free up the former for more general computing tasks, while also uplifting the fluency of 1080p, 4K, 4K HDR video playback. The higher your video resolution is, the better 5KPlayer hardware acceleration's CPU-saving efficiency will be.

To make it clearer, you can see the obvious data change, say CPU consumption saving rate in this statistics analysis table when switching from software decoding to hardware acceleration. When 5KPlayer hardware acceleration is enabled, CPU consumption drops sharply saving around two-thirds CPU usage especially on the quite old computer PC-1. In particular, 4K HDR is also playable and viewable on PC-1 with GPU acceleration available and CPU consumption also declines 16%-18% on PC-2 with better configuration.

Besides the CPU saving, another point should be mentioned is the GPU rendering effect after turning on hardware acceleration. Is the output image quality worse than that processed by the software decoding? From the experiment above, it is proved that there is little difference on the rendering effect between software and hardware acceleration that our human eyes can't distinguish, the output image quality including the hue, saturation and luminance looks great on 4K UHD monitor.

Now that there is nearly no influence, why not choose hardware acceleration to largely speed up the GPU rendering and save GPU render usage around 10% in average compared to software decoding?

3: Image Quality Comparison of Software Decoding and Hardware Accelerated Decoding

It is widely and theoretically believed that software decoding is distinctly better than hardware decoding on output image quality thanks to the CPU's simultaneous post-processing ability and the seamless applying of FFmpeg filters, however, the computer's GPU hardware decoding decodes, renders and sends the video file data to monitor without further processing.

However, the 3 figures below created in the experiment part-2 suggests an opposite fact against the theory of software-decoding-better-than-hardware-decoding.

Fig.1 1080p sample-1 on PC-2 software decoding vs hardware decoding

Fig.2 4K sample-2 on PC-2 software decoding vs hardware decoding.

Fig.3 4K HDR sample-3 on PC-2 software decoding vs hardware decoding.

The experiment suggests that at least from human's eye view, 5KPlayer delivers image output in a manner that there's little discernable difference between the hardware acceleration mode and the software acceleration mode on a 4K UHD monitor, with the only slightly discernable color tone change.

Conclusion

To sum up, 5KPlayer can play HD 1080p 4K high-res videos by default but the 4K HDR content could not be displayed on PC-1 without turning on 5KPlayer hardware acceleration. On high-end computer PC-2, all the video contents can be played without choke, while the difference is that 5KPlayer hardware acceleration helps decode and render videos without occupying much CPU. CPU consumption in processing high definition videos drops largely to enable more tasks and save battery life, and also quicken the GPU render without sacrificing the image quality. So, that's why we recommend enabling hardware acceleration when playing HD 1080p, 4K, 4K HDR videos.

5KPlayer Hardware Acceleration Lab