Undoubtedly, it has become a consensus in the industry that computing power is productivity. In particular, computing power, as a key productivity factor in the digital economy era, has become the core support and driving force for mining the value of data elements and promoting the development of the digital economy.
In this process, the digital economy driven by computing power not only appeared as an independent economic form of the information industry, but also penetrated into thousands of industries in the form of computing power. While driving the transformation and upgrading of the traditional economy and optimizing efficiency, It has also spawned a brand-new industrial system, which has become the key infrastructure supporting the in-depth development of the digital economy.
But at the same time, with the advancement of the new generation of information technology represented by artificial intelligence, human society has entered the era of wisdom at an accelerated rate, and traditional computing is also being upgraded to smart computing, and computing power is gradually upgraded to intelligent computing power. The emergence and development of intelligent computing power has provided a quantifiable technical force for innovation. As the saying goes, "as much computing power, there is as much productivity, and as much intelligent computing power, there is as much innovation power."
This is also an important reason why Inspur Information proposed "Intelligent Computing Creativity Integration of Data and Reality" at IPF 2022. Of course, the upgrade from computing power to intelligent computing power, and the transformation of productivity into innovation, is not achieved overnight. Based on this, Inspur Information has also comprehensively upgraded the digital-real fusion infrastructure framework to accelerate the construction of an innovative system of computing power, algorithms, data, and interconnection for the smart era. An intelligent computing base that integrates numbers and reality.
From this point of view, under the background that the country proposes to accelerate the implementation of the "new infrastructure" and "counting from the east to the west" projects, as well as the development of global intelligence and the accelerated arrival of the metaverse industrialization process, the "smart infrastructure" proposed by Inspur Information this time The brand-new concept of "integration of computing, innovation, digital and reality" and a series of key measures behind it to build a solid foundation for smart computing are not only forward-looking, but also an important manifestation of a holistic view.
The transition from computing power to intelligent computing power
In fact, computing power has become an important indicator to measure the development level of a country or even a region's digital economy. According to the "2021-2022 Global Computing Power Index Evaluation Report" jointly completed by Inspur Information, IDC and Tsinghua University Global Industry Research Institute, the computing power index and the trend of GDP show a significant positive correlation. From 2016 to 2020, for every 1 point increase in the computing power index of the fifteen key countries, the country's digital economy and GDP will grow by 3.5‰ and 1.8‰ respectively, and this trend is expected to continue in 2021-2025.
Not only that, when a country's computing power index reaches more than 40 points, every time the country's computing power index increases by 1 point, its driving force for GDP growth will increase by 1.5 times, and when the computing power index reaches 60 points or more, Every time a country's computing power index increases by 1 point, its driving force for GDP growth will increase to 3.0 times.
Peng Zhen, CEO of Inspur Information
This also means that computing power itself is productivity, and it has indeed created immeasurable unique value behind it, especially when the continuous combination of data, artificial intelligence and even machine learning, more and more industry users are also accelerating Intelligent computing power investment represented by AI computing power. The most typical example is in the fields of intelligent customer service and risk management and control in the financial field. Intelligent computing power greatly saves the cost of artificial seats and improves the work efficiency of enterprise customers; in the manufacturing industry, intelligent computing power also promotes the realization of intelligent manufacturing and accelerates digital The construction of factories; in the field of scientific research, intelligent computing power is assisting drug discovery, for the study of protein fold structure, the detection and identification of antibacterial resistance genes, and the development of new drugs, while greatly shortening the time for new drug development and marketing, etc. .
The evaluation from the "2021-2022 Global Computing Power Index Evaluation Report" found that industries with large investment in computing power also invest in the application of new technologies, among which the Internet, finance and manufacturing are among computing power investment and AI investment Average rank. At the same time, at present, the penetration of intelligent computing power into traditional industries such as transportation, industry, agriculture, medical care, and finance is accelerating, and the application scenarios are further expanded from general scenarios to industry scenarios, which not only leads to the acceleration of traditional computing power to intelligent The upgrading and evolution of computing power has made countries all over the world accelerate the process of upgrading the computing power industry.
For example, China's "14th Five-Year" Digital Economy Development Plan clearly mentioned that the digital economy is the main economic form after the agricultural economy and the industrial economy. Statistics show that in 2020, the added value of my country's digital economy core industries will account for The proportion of gross domestic product (GDP) reached 7.8%, and by 2025, this proportion is expected to reach about 10%. In addition, the large-scale "new infrastructure" in China and the recently launched "Digital from East to West" project are also enabling the in-depth application of digital technologies such as cloud computing, big data, and artificial intelligence in various traditional industries, promoting the transformation of various industries in production methods, business Profound changes have taken place in the model and management paradigm.
In the opinion of Peng Zhen, CEO of Inspur Information: "In the past few years, it has become the consensus of the whole society that computing power is productivity, and the country is also turning computing power into the driving force for the development of the digital economy through new infrastructure. Speaking of the huge changes brought about by the productivity of the whole society. And the digital transformation of the whole society has driven the upgrade of traditional computing power. Therefore, in the world depicted by digitalization in the future, intelligent computing power will become extremely critical. Only intelligent computing power can achieve innovation, and the key behind this is the intelligent computing center. Its core is to use algorithms, data, and intelligent computing power to transform productivity, turn productivity into the innovation power of the whole society, and realize the conversion of new and old kinetic energy. As a result, China's entire economic and social development will also step onto a new level."
Four new challenges behind the interpretation
It can be seen that when computing power is transformed into intelligent computing power, the "three elements" of productivity, laborers, labor tools, and labor objects, have also undergone significant changes. Among them, laborers have changed from simple people to "human + AI"; the objects of labor have extended from natural substances such as minerals and crops to data, and "material + data" has become the object of labor in the age of wisdom; the original machinery, The integration of labor tools such as motors with smart computing has promoted the intelligent upgrade of production tools.
In China, after years of continuous accumulation, the investment of computing power resources in thousands of industries and the organic combination of massive data resources, huge application demands, and an open market environment have formed the foundation for the development of my country's intelligent computing power. However, it also poses greater challenges to computing power, algorithms, energy consumption, and ecology.
As Peng Zhen said: "The fusion of numbers and reality is a process of gradual evolution. We will face more complex scenarios, greater amounts of information, and more efficient decision-making needs. Power, algorithm, data, network, promote the integration of data and reality, promote quality change, efficiency improvement, and the conversion of old and new kinetic energy." This also means that the new infrastructure behind the intelligent computing power also needs to be upgraded or even reconstructed, specifically look:
First of all, in terms of computing power , in the era of wisdom, the scale of computing power demand continues to expand, and the contradiction between computing power supply and demand is becoming more and more prominent. It can be seen that the data shows that the computing power of global data centers is growing on a large scale. It is expected that by 2020 to In 2025, the demand for computing power will increase by 30 times; the computing power demand for AI models will double on average in 3.4 months; in addition, even for a single application scenario, the computing power demand will become more and more massive, such as reaching the metaverse real-time The immersive rendering effect requires 1 million times the computing power of the current level. However, the current computer architecture is not suitable for the interconnection and integration of multiple computing power. To develop smart computing, it is necessary to innovate the computing architecture to solve the challenge of multi-computing power integration.
Secondly, in terms of algorithms , the entire algorithm is also facing the challenges of diversification, specialization, and massive quantification. Diversity is reflected in the fact that there are many types of large algorithm models and different capabilities and characteristics. In addition, there are nearly 1,500 modeling methods, nearly 100,000 algorithm models, and more than 40+ basic structures and 500+ different algorithms in the market. The technical complexity of the basic operator is unbearable for most enterprises; the professionalism means that the average construction time of the current algorithm application is 3 months, and the algorithm needs to be iterated quickly. In addition, new AI algorithms and new theories emerge in an endless stream, which makes The difficulty of cultivating intelligent computing talents has always been very high; giant quantification means that the model scale of the algorithm is getting larger and larger, and the large model with trillion parameters is close at hand. In this case, it is necessary to develop intelligent computing infrastructure, including computing power infrastructure and algorithm infrastructure, to lower the AI application threshold of traditional enterprises, including costs, talents, etc., to accelerate the process of industrial AI.
Third, in terms of energy consumption , as the world and China promote the "dual carbon" strategy, it is becoming more and more urgent to resolve the energy consumption challenge in the construction of intelligent computing centers. The data shows that the total power of current servers is continuously increasing. Taking China's server market as an example, it has increased from 690 million kilowatts in 2012 to 5.62 million kilowatts in 2020. This also leads to continuous power consumption of data centers. From 64.5 billion kwh in 2012 to 204.5 billion kwh in 2020, it is estimated that the energy consumption of data centers will reach 260 billion kwh in 2023, and the electricity consumption of data centers will exceed 400 billion kwh by 2030, accounting for The proportion of electricity consumption in the whole society will rise to 3.7%. More importantly, the current PUE value of the data center is still relatively high, with an average PUE value of 2.46, which is far below the national PUE value of 1.4 or below. Therefore, reducing energy consumption of the data center and following the path of green innovation in the data center Road will be the only choice in the future.
Finally, in terms of ecology , the entire market is currently facing the challenges of vertically closed systems and complex and discrete ecology, including repeated system construction, complete but not strong, and "various" chip architectures, development libraries bound to chips, and poor interoperability of AI frameworks And so on, and the different genres of cloud, big data, and AI technology developed on this basis are not compatible with each other, and AI technology is also incompatible with cloud, big data, and traditional applications. Only by relying more and more on applications and promoting the integration of different technical schools can we completely break the application bottleneck, realize the sound development of the intelligent computing industry, and realize the prosperity and vigorous development of the ecology.
It can be seen that at a critical moment when intelligent computing power has become the core driving force to promote China's digital economy and "new infrastructure" construction, how to solve the new challenges in computing power, algorithms, energy consumption and ecology is not only a "top priority". ", it is "urgent".
Build a solid foundation for intelligent computing that integrates numbers and reality
In this context, Inspur Information has also comprehensively upgraded the digital-real fusion infrastructure framework, accelerated the construction of an innovative system of computing power, algorithms, data, and interconnection for the smart era, and built a healthy development of the intelligent computing industry cycle, thereby better Build a solid foundation for intelligent computing that integrates numbers and reality.
First, continue to promote computing power and algorithm infrastructure, and realize the inclusiveness of intelligent computing centers. Peng Zhen said that computing power and algorithms are the core of the intelligent computing center. In order to generate first-class intelligence, first-class algorithms are necessary. At the same time, the efficient operation of algorithms depends on leading computing power, computing power infrastructure and algorithm infrastructure. Complement each other.
Therefore, in terms of computing power, Inspur Information will continue to consolidate its advantages at the level of computing power systems and cloud digital intelligence software platforms, and provide hardware-reconstructed smart computing complete machine systems, including general computing systems and smart computing acceleration systems, which will also be oriented to massive Data and high-speed interconnection provides intelligent computing storage and data center switch products. At the same time, Inspur Information will continue to extend the technology stack to the bottom layer, develop customized AI acceleration, network acceleration and storage acceleration products for different scenarios, and finally form a comprehensive system covering physical facilities, chips/accelerator cards, complete machine systems, and system software. Leading product system.
Among them, the release of MetaEngine, the industry's first Metaverse server, is the latest representative of Inspur Information's innovation in the technical field. MetaEngine can provide powerful computing power for the creation and operation of the Metaverse digital space. The goal is to build an integrated infrastructure of software and hardware. Serving large-scale, highly complex, high-fidelity digital scene collaborative creation and real-time rendering simulation, as a solid base connecting the real world and the digital world, accelerating the "digital and real integration" of various industries, and truly enabling all walks of life to "know the world, "Reproduce the world" has become more efficient and simple.
In the algorithm dimension, in order to conform to the development trend of algorithm infrastructure, Inspur Information has also formulated an evolution route for the upgrading and development of algorithm models. The first stage is mainly characterized by supervised learning and recognition applications; the second stage is single-mode, self-supervised The third stage is characterized by multi-modal and self-supervised learning; the next stage, the algorithm model will enter the super-modal active learning stage that interacts with the physical world, and currently Inspur Information is in the second stage In the process of evolution to the third stage.
In this regard, the Inspur Artificial Intelligence Industry Research Institute released the world’s largest Chinese artificial intelligence giant model “Yuan” last year, and at IPF2022, Inspur Information released the latest skill model generated by the “Yuan” large model, including Knowledge-enhanced dialogue models, knowledge retrieval question-answer models, Chinese-English translation models, and ancient Chinese comprehension models. The "source" large model can produce professional skill models more quickly, so as to realize miniaturized and lightweight landing operations in more professional scenarios. At present, the "Yuan" large model has also settled in the Nanjing Intelligent Computing Center. At the same time, within less than 5 months of its launch, more than 200 units have tried it out, supporting more than 100 industry applications.
The second is to firmly invest in the R&D and innovation of liquid cooling technology, and enable green computing power through the full-stack layout of liquid cooling. At present, Inspur Information has achieved large-scale deployment in various types of products such as cold plate liquid cooling, heat pipe liquid cooling, and immersion liquid cooling. At the same time, Tianchi, the largest liquid cooling data center research and development and production base in Asia, Publicly unveiled, Tianchi Base has built a full-chain liquid-cooled intelligent manufacturing capability from R&D, testing, production, and delivery, with an annual output of up to 100,000 units. The PUE value of the user data center is reduced to below 1.1, and the overall delivery cycle is only within 5-7 days.
It is worth mentioning that the Tianshu (ORS3000S) liquid-cooled complete cabinet server jointly released by Inspur Information and JD Cloud has also been deployed on a large scale in the JD Cloud data center. The cooling efficiency of the Tianshu liquid-cooled complete cabinet system It has increased by 50%, and the power supply energy efficiency has increased by 10%. At the same time, it can also be delivered in an integrated manner, with efficient operation and maintenance deployment, and the delivery efficiency has also increased by 5-10 times. It guarantees basic computing power, helps JD Cloud data centers reduce PUE, and lays the foundation for low-carbon energy saving.
In addition, Inspur Information also released the "Qiantang River" liquid-cooled intelligent computing center to upgrade the computing power infrastructure plan. The "Qiantang River" liquid-cooled intelligent computing center is an overall solution for a green intelligent computing center created by Inspur Information and Suiyuan Technology. Using Inspur Information's advanced complete set of liquid cooling technology, the overall PUE value of the intelligent computing center has also dropped below 1.1.
"Full-stack liquid cooling has become a very important strategy for Inspur in the green and low-carbon era. At present, Inspur is cooperating with many industry customers to discuss how to formulate standards for the entire liquid cooling technology in the future and how to develop the industrialization ecology of liquid cooling. At the same time, we also hope to apply for more liquid cooling patents through technological innovation, so as to form a more competitive position on the commanding heights of the entire liquid cooling industry in the future, and promote the long-term and sustainable development of the entire liquid cooling industry in China. Sustainable development." Peng Zhen said.
The third is to jointly build the Yuannao ecology, accelerate the construction of industry wisdom, and build a solid foundation for the integration of numbers and reality. In order to solve the challenges of complex and discrete ecology and difficult implementation of industrial AI, Inspur Information led the launch of the Yuannao Ecological Plan in 2019, bringing together left-hand partners with core capabilities in AI development and right-hand partners with overall industry AI solution delivery capabilities. In 2022, Inspur Information will continue to implement the "Intelligent Computing Partner" plan to continue to expand the Yuannao ecosystem, specifically including upgrading the Yuannao platform AIStore, accelerating the transformation of smart distribution, co-creating industry smart solutions, and empowering gas stations with partners, etc. Four major measures. To this end, Inspur Information will train more than 5,000 technical experts for its partners and increase investment in the development of industry smart computing solutions.
In this regard, Peng Zhen finally emphasized: "In the development of the entire digital economy and the intelligent transformation of all walks of life, Yuannao Ecology will play a very important role, because the innovation of computing power algorithms is based on technological innovation. In order to transform them into intelligent computing power and creativity, the whole society must do collaborative innovation, and Yuannao ecology will also provide everyone with a richer soil for innovation and fertile soil for development, and in the future, the wave will also Guided by intelligent computing, we will continue to deepen the Yuannao ecology, and respond to the development opportunities of digital and real integration with innovation."
In general, if the entire society has completed the process of informatization in the past two decades; then in the next ten years, the entire society will accelerate from informatization to digitalization, and will make great strides into the age of wisdom. It was a difficult climb to the top, coupled with the accelerated emergence of new technologies, the implementation of policies such as the "Double Carbon" strategy and the "East Count West Count" project, while providing more innovation possibilities for the entire smart computing industry, It also brings more transformation and upgrading pressure, and Inspur Information promotes the inclusiveness of the new infrastructure of the intelligent computing center through continuous innovation of computing power, algorithms, and data infrastructure, and will definitely become an important force driving the high-quality development of the digital economy , to better build a foundation for intelligent computing that integrates numbers and reality, and its value can be said to be "not limited to the present, but also related to the future."
Shenyao's Science and Technology Observation was founded by Shenji, a self-media person who crosses the border between technology and automobiles. With 18 years of experience in enterprise-level technology media work, he has long focused on the observation and thinking of enterprise digitalization, industrial intelligence, ICT infrastructure, and automotive technology content.
