Inspur released a new AI brand TensorServer, and set itself a small goal of "world first"

640?wx_fmt=gif&wxfrom=5&wx_lazy=1


The article is reproduced with permission from Raicworld (ID: raicworld)


At the Inspur IPF Conference in 2015, Inspur announced that it would enter the top three in the world within 5 years.


Two years later, according to Gartner data, Inspur's server shipments rose to third in the world in 2017, becoming the world's major server solution provider, completing the transformation from China's number one to the world's leading manufacturer three years ahead of schedule.


On April 26, at the IPF2018 Inspur Partner Conference, Wang Endong, an academician of the Chinese Academy of Engineering and the executive president of Inspur Group, announced Inspur's next five-year "small" goal: focus on intelligent computing and achieve the world's number one server.



640?wx_fmt=jpeg&wxfrom=5&wx_lazy=1



We are accelerating into the era of wisdom, and intelligent computing will become mainstream


With the global upgrade from the traditional information society to the intelligent society, we are accelerating into the era of wisdom. Each era has key scenarios that support its characteristics. In the era of wisdom, cloud computing, Internet of Things, big data, Five scenarios including artificial intelligence and industrial Internet. These scenarios are closely related to computing, and smart computing will become the mainstream computing form in the future.


What is smart computing?


640?wx_fmt=jpeg

Wang Endong, Academician of the Chinese Academy of Engineering and Executive President of Inspur Group


Wang Endong, academician of the Chinese Academy of Engineering and executive president of Inspur Group, said that smart computing is an innovative computing form that obtains information and knowledge from data, with cloud computing as the basic platform, big data as the cognitive method, and deep learning (Deep learning) is an optimization method, referred to as CBD.


In fact, in 2015, when IPF Inspur released the "Computing +" strategy, and smart computing was first mentioned, Inspur has been practicing smart computing. Nowadays, the arrival of the era of wisdom is both an opportunity and a challenge for Inspur. In order to keep up with the development trend of the times, Inspur will continue to innovate in product technology, business model, manufacturing and other fields to seize the Industrial opportunities for intelligent transformation. As for the development in the next five years, Inspur has a clear goal: to establish a global R&D system, supply system, and service system, and to achieve the world's first server.


The new AI brand "TensorServer" was launched, carrying and enabling the release of AI value


Nowadays, with the rapid development of AI, more and more enterprises are beginning to enter AI. But for traditional enterprises, the threshold of AI is still very high. In mid-September last year, Inspur and Baidu released the ABC all-in-one machine, which is a software-hardware integrated AI application solution deeply customized for industry users. It adopts Baidu's DL framework, mature algorithm model and cloud management technology, as well as Inspur's The AI ​​computing hardware platform covers the requirements of model training training and online reasoning inference. It can be used out of the box and is a turnkey solution. The purpose is to lower the threshold for traditional enterprises to enter the AI ​​process and empower traditional enterprises. And now, Inspur is still cultivating on the road of AI empowerment.


Inspur has always attached great importance to product innovation. As Wang Endong said, Inspur servers have a very sufficient product layout for cloud computing needs. It is the only manufacturer in the world with four open standard server products. Facing the field of deep learning, Inspur has the world's most powerful AI server with the highest density. And with the most complete AI computing products in the industry, combined with AIStation, Inspur can provide a complete solution for deep learning clusters.


At this year's IPF2018, Inspur released a new AI brand TensorServer. Tensor is the basic element of AI algorithm, and Server is the infrastructure of computing power. TensorServer is intended to be the carrier and enabler of AI, giving Inspur AI a new connotation.


640?wx_fmt=jpeg

Inspur releases new brand of AI - TensorServer


Specifically, TensorServer integrates and emphasizes Inspur's powerful AI infrastructure capabilities, and can deliver AI infrastructure systems with strong self-adaptation, efficient acceleration, and flexible expansion, covering four levels of platform, management, framework, and application, through AI computing Full-stack AI end-to-end infrastructure layer products and solutions such as continuous innovation and research and development of hardware platforms, efficient implementation of AI system management processes, continuous optimization of AI frameworks, and acceleration of AI industry applications.


Through the core value and capabilities of a series of AI computing power provided by TensorServer, enterprises are empowered to have solid and reliable infrastructure support in the era of artificial intelligence, fully release the value of artificial intelligence, and help them take the lead in the fierce competition.


640?wx_fmt=gif

Inspur and iFLYTEK jointly release AI Booster


At this conference, Inspur and iFLYTEK released the "AI Booster" cluster for offline training scenarios of speech recognition and intelligent translation, and released the first elastic GPU server NF5468M5 for AI cloud scenarios. NF5468M5 mainly has the following characteristics:


It can flexibly support offline training and online inference scenarios of AI models, and meet the performance requirements of offline training. It can support 8 highest-performance Tesla Volta GPUs with high-speed interconnection of PCI-E or NVLink; for online inference For high energy efficiency, it can support 16 Tesla P4 GPUs with high energy efficiency, suitable for voice, picture and video scenarios. In the intelligent video analysis scenario, it can process more than 300 channels of 1080p high-definition video structures at the same time.


The elastic configuration capability optimized for AI cloud scenarios enhances the flexibility of PCI-E topology and quantity ratio, adapts to the computing requirements of various frameworks, and adds remote GPU topology reconstruction technology that is easy to manage and maintain.


·Maximize data storage and communication performance and eliminate IO bottlenecks. The NF5468M5 can support 288TB mass storage or 32TB solid-state storage, achieving up to 400Gbps communication bandwidth and ultra-low latency of 1us.


In the past 2017, Inspur AI servers have grown by more than 600% year-on-year, which is 4.5 times the average growth rate of the industry. Its AI market share in China exceeds 60%, and the AI ​​servers of Alibaba, Tencent and Baidu are 90%. The above is from Tide.


Drawing on past experience, Inspur is ready to launch a "five-year plan" for the future


In 2017, Inspur's server shipments have risen to the third in the world, and completed the transformation from the first in China to the top three in the world three years ahead of schedule. The reason why this goal can be accomplished ahead of schedule can be attributed to the four core driving forces behind Inspur. First, product innovation, focusing on the development of cloud data center-oriented converged architecture cloud servers, including the whole cabinet system SR, InCloudRack and The blade system I9000, etc., was the first to develop the SR of the whole cabinet server, and the global shipment of cloud servers ranked first in the third quarter of 2017; second, seize the opportunity of the rapid development of the Internet, innovate the business model, form the JDM collaborative innovation model, and win the global Internet operation Thirdly, Inspur’s own intelligent manufacturing production line realizes the comprehensive upgrade and transformation from traditional manufacturing to large-scale intelligent manufacturing, creating a new industrial form of mass customization of complex information technology products; fourthly, it embraces open source at the same time Technology, active in international open source organizations such as OCP, ODCC, Open19, OpenStack, etc., and better integrated into the latest global technology ecosystem.


640?wx_fmt=jpeg

Peng Zhen, Vice President of Inspur Group


Peng Zhen, vice president of Inspur Group, said: Inspur's innovation in the field of converged architecture is not only in product and strategy, but also in business model innovation. We redefine the business model of JDM, cooperate with customers in R&D, joint production, and respond quickly to market demands. This agile innovation model is the core essence of JDM and an important capability model for Inspur to win the market in the Internet age.


For the next development of Inspur AI, Inspur has launched a new AI ecological development plan code-named T, which aims to release the core value and capabilities of AI computing power from the three dimensions of innovation, partners and talents, and empower enterprises in The era of artificial intelligence has solid and reliable infrastructure support.


Inspur will rely on the breadth of applications from the Internet to industrial AI scenarios and the deep accumulation of continuous investment to enable partners to extensively participate in the commercialization of AI. Inspur will help partners to innovate AI R&D, customer development, and share AI power, and form a shared and collaborative coupling with partners in terms of establishing R&D capabilities, improving business models, and participating in industry ecology, so as to jointly create a healthy AI business ecosystem.


As for talent cultivation, Inspur will launch a large-scale AI computing talent training program, unite with the world's largest supercomputing competition, and focus on cultivating college students in related fields; establish an open cooperation community for AI computing, and build an international advanced exchange platform AICC; establish Inspur AI computing youth Talent Award, set up tens of millions of funds to encourage young talents to break through the cutting-edge cutting-edge.


In the future, in the face of new opportunities and challenges, how will Inspur achieve the "small" goal of becoming the world's No. 1 in the past five years? And let's wait and see.


END

The above article is reproduced and does not represent the opinion of this newspaper


640?wx_fmt=jpeg

Guess you like

Origin http://43.154.161.224:23101/article/api/json?id=326023504&siteId=291194637