Who is promoting the new inflection point of audio and video industry in 2022?

​

be based on XPU Hardware strategy for 、 Open software stack , Intel is pushing the audio and video industry to dynamic 、 Enrich 、 The next inflection point of multi dimension and interactivity

author | Fight

Produce | Industrialist

2022 It was called in “ Cloud show ” First year .

Hundreds of millions of people poured into Jay Chou's online concert , To westlife Customize online concerts for Chinese fans and brush the screen in the circle of friends .“ Cloud show ” Remove people from the haze of the epidemic , Instantly pull to the noisy scene .

People are recalling their youth 、 While remembering the past , It is also an audio-visual feast brought by the performance on the exclamation line .

In the past , SMS 、 The website is the only way to interact with stars ; Later, live broadcasting rose , Play online for idols call, It has narrowed the distance between fans and stars , But in the delay time 、 The Caton frequency is not satisfactory . for example , In the live broadcast of Sports Events , There are often match point jams , When it returns to normal , But I have already missed the wonderful moment ; Now , With the optimization and upgrading of live broadcast technology , These have undergone fundamental changes , Instant interaction 、 Video fluency has become standard .

Behind the rapid development of the live broadcasting industry , Also, audio and video manufacturers continue to empower with technology and services , So as to drive the upward development of the industry .

Under the epidemic , Limited travel , Online performance 、 Online office 、 Online recruitment 、 Live broadcasting with goods has become the norm ; in addition , Under the concept of meta universe ,VR、AR The field is rising again ; In addition, the short video platform has a strong growth trend . More and more application scenarios are mined .

According to the IDC released 《 China Video cloud market tracking (2021 In the first half of the year )》 Show ,2021 In the first half of the year, the scale of China's video cloud market reached 43.7 Billion dollars , Year on year growth 38.7%.

Beyond all doubt , The audio and video track is facing unprecedented development opportunities .

However , New opportunities also correspond to new challenges . The audio and video track standing at the tuyere , Who will be the next breaker ？

One 、 Under the wind , Reproduce the industrial dilemma

1872 year , British photographer Mike bridge , In order to test whether the horse will take off in the process of running , With the help of 24 A camera , Continuous shooting 24 A picture , This created the first in human history “ video ”.1877 year , The famous inventor Edison was in his “ Cylinder Phonograph ” On , Recorded what he read himself 《 Mary has a little lamb 》 The lyrics of . The sound of just a few seconds , It has become the first audio in human history .

thus , Mankind has officially entered the era of audio and video .

Technology changes , Audio and video from film to disc , Then turn to the mobile phone 、 On mobile devices such as computers , The load-bearing methods are constantly changing . Get into 21 After a century , The explosion of short video services , And the rise of nationwide live broadcasting , It has pushed the popularization of audio and video to a new climax .

During this time ,“ Traffic ” Become an indicator that users pay more attention to . The Internet audio and video platform and its underlying technology , Whether it's consumers 、 Enterprises or investors , Not paying too much attention to it .

However , Outbreak of new crown epidemic , Breaking this inherent situation .

Under the catalysis of the epidemic , Business activities and population movements are blocked , This has led to the instant release of strong demand for remote collaboration and online entertainment .

for example , Offline activities of enterprises are forced to migrate to online activities 、 A large number of enterprises adopt the mode of remote office instead of offline office 、 The school opens “ cloud ” Classroom ...... The demand for real-time interaction has soared . According to the consulting firm IDC Data disclosure ,2020 Throughout the year , Compared with the same period last year, the scale of China's video conference market has increased 18.9%, Reach about 65.2 RMB 100 million , Explosive growth .

in addition , The market scale of online education scene has also ushered in an epoch-making growth trend , Relevant data display ,2020 In, China's education real-time audio and video market was relatively large 2019 Year-on-year growth 46.9%, achieve 47 One hundred million yuan .

The application of audio and video is more and more extensive . And then stimulate the rapid growth of market scale . Specific data shows ,2020 In, the market scale of China's audio and video solutions broke through 400 Billion , The annual growth rate exceeds 55.1%.

Under the dual effects of technological innovation and market environment , Audio and video have become the Internet “ coal 、 water 、 electric ” As important as , It has gradually become one of the new infrastructure at the bottom of the Internet industry .

However , With the development of the audio and video industry , The penetration rate of audio and video systems in various application scenarios is approaching saturation , In addition, users' pursuit of ultimate experience , This makes each platform have more diversified requirements on the technical level , The challenges faced by the underlying computing power are also increasing .

Take the live broadcast , With big data 、 The rise of emerging technologies such as cloud computing , as well as 5G And the rapid development of communication technology , Live streaming to the cloud has become a trend , How to connect the audio and video system with the cloud platform becomes a new demand .

secondly , With the rapid penetration of various high-definition audio-visual equipment , Technical solutions are needed to deal with higher definition faster 、 Video codec and transcoding with lower delay . To deal with 1080P、2K Even 4K Continuous under video “ Capacity expansion ” Audio and video data stream .

in addition , At present, the direct broadcast ecology has become the general trend , Need to cooperate with e-commerce 、 More service platforms such as business are connected , This puts forward higher requirements for the underlying infrastructure carrying audio and video capabilities , These requirements include faster computing power 、 More flexible calculation force distribution scheme 、 Better data storage performance and stronger network throughput .

And now AI technology Continuously enable live broadcast , Such as product recommendation 、 Instant effects 、 Interest content information flow 、 Virtual image, etc , Match audio and video capabilities with AI Deep integration of Technology . How to provide efficient services for audio and video capabilities AI The framework and acceleration capability have also become the new requirements of the new live broadcast scene .

In short , With the rapid development of emerging technologies , Audio and video technologies need to be constantly upgraded to support new applications 、 Requirements under new patterns and scenarios .

The fact that , At present, for the whole audio and video track , It not only needs the optimization of the upper software system , It also needs to be on the underlying hardware infrastructure , Find a way to break the situation . Based on this , To realize the true subversion of audio and video track 、 Innovative technological upgrading .

Two 、“ Software optimization + Hardware acceleration ”, Improved anchoring performance

The key to breaking the game , Often lies in the underlying foundation .

2021 year 2 month 9 Japan 24 when , The New Year bell rings on time , For the Chinese people , The whole family sat around the TV , Watching the live TV broadcast of the Spring Festival Gala together is a standard configuration for every new year .

However , What's different is that , Because of the impact of the epidemic , This year's Spring Festival Gala first adopted “ Cloud live ” In the form of , High concurrency video stream processing is required , Upload the video being recorded to ECs , After processing, it is distributed to a large number of user terminals .

Based on this , In a live broadcast , How to implement efficient transcoding on the cloud 、 Reduce delay and improve image quality , Become a challenge faced by major cloud manufacturers , It is necessary to continuously optimize and upgrade the underlying architecture and technology of its own platform .

Take Jinshan cloud as an example , In order to improve the image quality 、 While reducing bandwidth pressure , Improve the performance of cloud transcoding , Achieve higher performance density , Its presence “ Smart HD ” The ECS of is equipped with Intel To the strong processor .

among , The processor is customized and optimized according to the characteristics of the public cloud application load , Provides up to 32 A physical core , The operating frequency reaches 2.6GHz. In single core performance 、 The number of cores has been significantly improved , by “ Smart HD ” The cloud transcoding performance of the system has laid a solid foundation .

in addition , The application scenarios of Jizhi HD have strict requirements for delay , So Kingsoft cloud uses intel AVX-512 To optimize convolution overloading . Compared with the traditional convolution implementation , This method has the advantage that there is no need to expand the edge of the source image 、 No need to fill the filter 、 There is no need to transfer the entire filter buffer 、 No need to rotate the source image / Filter / Output and other advantages , It can effectively reduce the delay and bandwidth of online image processing .

At Intel AVX-512 With the support of instruction set , Jin Shan Yun ERJND The module can realize 48-103 Times performance improvement .

besides , Intel The integrated performance primitives also provide performance optimization for the new solution .

Intel IPP It can quickly realize the discrete cosine transform in a function call （DCT）, While improving the operation efficiency , Greatly reduced the amount of code written . By using Intel IPP Library function 、 Optimize algorithms to minimize memory footprint 、 Intel AVX-512 Instruction set optimization, etc , Jinshanyun is in DCT Computing, etc 3 Times the performance improvement .

Before and after optimization DCT Performance comparison

in addition , Jinshan cloud also relies on Intel To the strong Processor integrated Intel Deep learning speeds up （ Intel DL Boost） technology , take “ Smart HD ” The numerical accuracy in the application of deep learning is FP32 The model is converted to use VNNI Instruction set optimized INT8 Numerical accuracy .

use INT8 Lower precision values such as can make better use of the cache , Increase memory data transmission efficiency , Reduce bandwidth bottlenecks , Thus, computing and storage resources can be more fully utilized , And reduce the system power .

It means , With the support of the same resources ,INT8 It can bring more operations per second for deep learning reasoning （Operations Per Second,OPS）. In this way , On the premise that the accuracy meets the requirements, Jinshan cloud , Greatly improve the performance of deep learning .

With the help of Intel Software and hardware products and technologies , Jinshan cloud video solution has achieved significant bandwidth savings 、 Better video experience 、 Picture quality is improved 、 Lower delay and many gains .

The Intel hardware infrastructure is like a “ Performance accelerator ”, Through Intel To the strong Scalable processor 、 Intel FPGA product 、 Intel Haughty Hardware product portfolio such as persistent memory , Provide powerful computing for all kinds of innovative solutions based on audio and video 、 Storage and network processing capabilities .

Software is more like “ System optimizer ”, In different application scenarios, a complete software stack is used to accelerate the performance of audio and video capabilities .

The combination of hard and soft , Intel for all platforms Provide audio and video processing capability based on cloud or data center 、 Professional and efficient codec / Transcoding capability , Enabling cloud service providers to bring more fluency to users 、 Higher definition 、 Simpler interaction 、 More affordable audio-visual experience . meanwhile , It also enables cloud vendors to promote system optimization more effectively , Reduce operating costs . It's finally done “1+1>2” Synergy of .

3、 ... and 、 New scenes generate new values , Moving towards the industrial inflection point

In the new scene , The value of audio and video industry has been redefined .

With audio and video products 、 Technology upgrade iteration , Constantly promoting new models 、 The development of new formats . At the same time, it also brings more imagination space for new application scenarios .

2021 In the second half of , The concept of meta universe is hot out of the circle , Highly recognized by the capital circle . among , Immerse 、 Interaction 、 The real-time full truth audio and video has undoubtedly become one of the important presentation forms of the meta universe .

The short term , The popularity of the yuan universe promotes audio and video VR Release of technical requirements , Achieved small-scale growth in the audio and video solutions market . But from the perspective of long-term development , Yuancosmos is more like an audio and video track , A microcosm of the new application scenario . Intel is helping to continuously upgrade audio and video technologies , It provides sufficient for the application scenarios of the meta universe “ Nutrients ”.

at present , In the cloud game scene with strong technical overlap with the metauniverse , The rapid development of market scale and the continuous enrichment of game categories , Bring more challenges to the cloud game platform , Need to have strong computing power and graphics processing ability , To provide low latency 、 Ultimate experience with high image quality , At the same time, we also need to choose more cost-effective software and hardware products to build a basic capability platform .

As a continuous innovator among smart device manufacturers ,OPPO Combine actual business and infrastructure , Launched with Intel The server GPU 、 Intel To the strong Scalable processor , And Intel Android Cloud Gaming Software Stack（ACGSS） Software stack OPPO Cloud game platform , Achieved significant performance improvements , In order to ensure the fluency of users 、 Stable gaming experience while , Also effectively controlled TCO.

OPPO Cloud game platform

from XPU Hardware strategy to open software stack , Intel through Intel To the strong Scalable processor 、 Intel FPGA product 、 Intel Haughty Persistent memory , And Intel Ethernet network adapter 、 Intel Visual cloud media analysis accelerator card and other hardware products , Provide powerful computing for innovative solutions based on audio and video capabilities in various scenarios 、 Storage and network processing capabilities ; In software optimization acceleration , Intel Media SDK、 SVT、 Intel oneAPI etc. , can It helps to accelerate the performance of audio and video capabilities with a complete software stack in different application scenarios , Help users quickly establish an end-to-end overall solution from data to application , And achieve better hardware performance .

With audio and video in the metauniverse 、 Smart home 、IoT、 Medical and other application scenarios are more extensive , Intel will be based on these products and end-to-end technology solutions integrating software and hardware , Continuously meet the needs of audio and video technology capabilities in various application scenarios , Fully reflect flexibility 、 reliable 、 Efficient architecture advantages . Keep focusing on new scenes , Empower “ A thousand lines and a hundred industries ”, So as to promote the audio and video industry to be dynamic 、 Enrich 、 The next inflection point of multi dimension and interactivity .

at present , stay AR/VR scene , Intel and Beijing Mobile , When it became popular, Qitian created 5G VR New experience of E-sports ; In the intelligent voice scene , Intel helps Tencent cloud microenterprises , Improved real-time speech synthesis performance ...... besides , Intel is constantly meeting the needs of more Internet applications for audio and video capabilities , Click on Read the original Learn more about audio and video innovation practice cases ~