In the era of data-centered innovation, the role and role of software are becoming increasingly important. With more than 15,000 software engineers and an ecosystem of more than 12 million developers, this is the "soft power" (soft power) of Intel, which is known for its "hard power" (hardware). Software and hardware collaboration and ecological innovation are Intel's strategic weapon for diversified computing in the future.
moves towards a data-centric era, and the computing architecture is expanding from CPU to XPU. Since the PC era, software has been a powerful gene flowing in Intel's "blood", but its huge value is still unknown. As Intel continues to deepen its data-centric transformation, software, as an important strategic asset, plays a more critical role in Intel's promotion of future computing innovation. For every order of magnitude performance enhancement potential of a brand new hardware architecture, software can bring more than two orders of magnitude performance improvements. Intel has unique advantages to unite important partners in the industry to jointly reduce development complexity under various architectures and workloads, and accelerate the large-scale deployment of innovative technologies.
Intel empowers partners and customers with unique ecological value, including providing simple and scalable tools to accelerate application development and deployment; adhering to customer first and optimizing for different types of workloads; adhering to the open concept, promoting open source and open ecosystem collaboration; connecting industry, academia and research, and cultivating future-oriented innovative talents. As the largest comprehensive R&D institution in China covering Intel's product departments, several experts from Intel Asia Pacific R&D Co., Ltd. recently shared Intel's ecology and value in the software field.
General Manager of Intel Asia Pacific R&D Co., Ltd.:
uses soft and hard collaboration to promote ecological win-win
As workloads such as AI, cloud data centers, Internet of Things, next-generation networks, and autonomous driving continue to emerge, people need to build different types of computing architectures such as GPU, FPGA, AI chips, and vision processing chips based on the core capabilities of the CPU to meet the needs of new workloads. Lu Ju, general manager of Intel Asia Pacific R&D Co., Ltd., said: "From PC-centric to data-centric, in the process of Intel's transformation, the potential market size has expanded to about US$300 billion. Therefore, there are more target customers and more problems to solve, and the circle of friends also needs to be larger. The ecosystem has always been a part of Intel's attention. In addition to the hardware ecosystem, the software ecosystem has also made great investment and cooperation. The entire computing architecture is not the original single CPU architecture. From CPU to XPU, software plays a key bridge role in it."
Intel Asia Pacific R&D Co., Ltd. was the predecessor of Intel Asia Pacific R&D Co., Ltd., which was established in Shanghai in 1993. "At present, our scale has more than 2,000 formal employees," Lu Ju introduced, "Intel Asia Pacific R&D Co., Ltd.'s corporate goal is to build a world-class engineering innovation center for Intel in China; our development strategy is to be rooted in China and look at the world."
In cloud computing , virtualization technology, big data, deep learning, basic input and output systems, firmware, video technology, server design and other fields, Intel Asia Pacific R&D Co., Ltd. has world-class engineering R&D capabilities. In the software field, Intel Asia Pacific R&D Co., Ltd. has in-depth and lasting efforts.
Feng Xiaoyan, chief open source scientist of Intel Asia Pacific R&D Co., Ltd.:
allows open source software-based solutions to have the best user experience
open source software has long been the mainstream software, and all levels of the system software stack have options for open source software, especially the major domestic cloud service providers basically build solutions based on open source software. Intel's growing software portfolio and extensive open source ecosystem projects confirm our emphasis on software and lead the industry to drive product and service innovation in a data-centric era.
"Intel has fully embraced open source software for nearly 20 years. Intel has established a team called the Open Source Software Technology Center," said Feng Xiaoyan, chief open source scientist of Intel Asia Pacific R&D Co., Ltd., "In January 2005, Intel then CEO made a promise: Intel launched a variety of driver software, and we will provide Linux solutions while providing Windows solutions, which marks the full embraced open source software."
Intel has a long history in open source and has contributed to open source all over the world. Intel has contributed to various open source projects and is also a member of various open source communities . Intel has always been an important contributor to a large number of open source community projects such as LinuxKernel, ChromiumOS, OpenStack, and is a major supporter of the Linux Foundation, Apache Foundation, Eclipse Foundation, and Open Source Software Promotion Alliance.
started with Intel to make open source software and has always been working (as shown below). Among them, Linux kernel and Kernel/KVM are the most important basic software for establishing various cloud services, and Intel ranks first in the community's code contributions all year round; Intel has also held consecutive China Linux kernel developers conferences with community partners since 2006. In terms of investment in open source software, Feng Xiaoyan gave an example, saying that Intel has promoted the active and widespread application of OpenStack in the Chinese community, and has a team dedicated to OpenStack in China; and has also established a team dedicated to CSP (cloud service provider) in China. "The purpose of our work in the open source software field is to hope that the entire open source software-based solution can run better on the Intel platform and have the best user experience," said Feng Xiaoyan.
Intel's leadership in the open source community has driven the industry, providing a new model for hardware and software interactions in emerging workloads.
Intel Architecture, Graphics and Software Department (IAGS), senior software architect Huang Shengsheng:
Accelerate unified data analysis and artificial intelligence technology innovation
In a data-centric world, more than half of the world's data has been generated in the past two years, but less than 2% of the data has been analyzed, and there is also a massive "silent data". One of the important reasons is that the process of new technologies such as artificial intelligence from the laboratory to the real production line lacks unified data analysis and artificial intelligence technology tools.
Intel Corporate Architecture, Graphics and Software Department (IAGS), senior software architect Huang Shengsheng believes that to solve such problems, we need to solve the gap in software and hardware architecture. Challenges faced by large-scale artificial intelligence applications include: the complexity of algorithms, the complexity of deployment, the complexity of data processing, as well as cost, whether it is scalable, proprietary interfaces, data privacy, etc.
Intel's two projects based on Spark development and open source: BigDL and AnlyticsZoo are bringing tangible benefits to customers.
Huang Shengsheng shared the end-to-end customer cases across industries. For example, by adopting Intel solutions, JD.com has improved 3.83 times compared to the original solution in terms of object detection and image feature extraction; the preprocessing performance of Midea Industrial Vision Detection Cloud Platform is 4 times higher and the inference performance is 16 times higher.
On the other hand, Intel is also continuing to promote ecological co-construction. In June this year, Intel established the Big Data Analysis and Artificial Intelligence Innovation Institute in China. The purpose of this is to promote Intel's extensive cooperation with the global ecosystem of big data and artificial intelligence for open source communities, academia, partners and end users.
, chief engineer of machine learning, Intel Asia Pacific R&D Co., Ltd.:
serves developers, allowing more companies to make good use of AI
. Intel not only provides convenient software tools for the industry, but also continuously promotes the development and optimization of software tools, ensuring that developers and ecosystems use software tools to benefit from hardware innovation.
Lin Xiaodong, chief machine learning engineer at Intel Asia Pacific R&D Co., Ltd., called himself " coder coder". "Many people use compilers, debuggers, tuners, programming languages and frameworks to develop tools. Our task is to optimize these compilers, frameworks, various libraries, etc., so that these developers are more efficient, serve developers, and build an ecosystem. The most important developers are developers targeting AI and deep learning."
"Intel's software is enabling (enable) and optim ization (optimization), what we do enables every transistor of all our hardware to fully utilize its capabilities. This is the purpose of Intel's software, which enables developers to make full use of our software and hardware very easily." Lin Xiaodong said, "Our team, from the beginning of being the world's software to the localization of China, now we are truly in China, creating software and serving the world according to China's specific needs."
He gave an example: from Broadwell to Skylake, the combination of software and hardware has brought up to 277 times the performance of machine learning. From Skylake to CascadeLake, the performance has increased by 28 times, of which it has been improved by about 4 times through hardware. The rest is all brought by software optimization, which includes some special optimizations made by Intel in the process of communicating with Chinese customers and understanding their workloads.
Intel continuously provides simple and scalable tools to accelerate application development and deployment. For example, oneAPI will simplify and unify programming across different architectures such as CPUs, GPUs, FPGAs, AI chips and other accelerators; the OpenVINO toolkit enables rapid development of high-performance computer vision and deep learning vision applications. He further pointed out: "Our team is also trying to solve a problem. In addition to serving big customers, we have observed that many traditional enterprises have encountered challenges. The only thing they have is: they know that AI and deep learning can definitely help solve problems. They have data, but they cannot find corresponding talents. In response to this, our local team will also make some innovations. We hope to help Chinese companies move towards the future without hiring developers or lacking relevant professional knowledge." Li Hong, Director of Intel's cloud computing software performance optimization team in China:
interacts with customers in both directions to achieve performance and optimization
In addition to having long-term open source contributions and rich software tools, Intel also has the ability to optimize customers' workloads, helping customers and partners accelerate the creation of applications and launch them to the market.
"Performance and Optimization" is not a new topic, but a classic topic that runs through various historical periods of computer technology development." Li Hong, director of the cloud computing software performance optimization team in Intel China, said, "In the process of interacting with customers, we are actually two-way. On the one hand, we will introduce some of Intel's technologies to customers' software development to help them improve performance. In this process, we have also deepened our understanding of their software, knowing how our technology is used in it, and what kind of technology they need in the future. We feed such information back to the product department to build a good foundation for future hardware or software products. Whether it is to customers or from customers to the backend, our focus is; performance and optimization, which is a hub connecting the two points."
He emphasized that Intel's investment in partners and software is long-term and continuous, "because I have been working in this team for 15 years. This team is a global team. We work with our partners on projects and engineering, and we also help them cultivate talents. This year, our partners in China have optimized different applications, about 70, and this number is expected to exceed 80 by the end of the year. The scope is very wide, including search engines, media processing, storage, etc., especially AI, and more than half of the projects are related to AI. Most programs can achieve a performance improvement of 2-4 times after optimization, and some can improve 10 times."
" Only by using both soft and hard can we better meet the needs of future computing.Through the cooperation of software teams and partners, Intel is committed to promoting the industry's software ecosystem forward. "Li Hong said.
University Cooperation Manager Yan Li, Intel Asia Pacific R&D Co., Ltd.:
Long-term investment, cultivating future innovative talents
talents is timely innovation. Intel continues to cooperate with universities for a long time and close cooperation to cultivate future-oriented innovative talents, including joint research projects, joint laboratories or joint innovation centers, joint curriculum development, teachers and student projects, such as participating in the Ministry of Education's "Industry-Academic Cooperation Collaborative Education Project" and carrying out series of competitions
"Intel has always been very passionate about education, which is related to Intel itself, and innovation is a core gene and is related to it. Education is a powerful tool for cultivating innovative talents and can help us hold the key to open the door to innovation. Intel has a deep understanding and rich experience in innovation and innovative talent training, so we have been systematically promoting this matter and making a systematic layout. "Yan Li, an university cooperation manager at Intel Asia Pacific R&D Co., Ltd., said.
. She emphasized: "We always tend to bring the latest technology to the academic community and share it with the school. There are many software and hardware platforms that we may not have been used by the industry and industry, so we will try them in school first. Moreover, "Intel has always tended to be a century-old store in university cooperation. Intel realizes that it is really a matter of ten years of growth, so long-term investment will lead to results. At the same time, our university cooperation is completely talent-oriented. Intel never sells any products to them. We just share technology and jointly cultivate talents with them. "
As data-centered transformation accelerates, workloads such as AI, cloud data centers, Internet of Things, next-generation networks, and autonomous driving continue to emerge. Intel will adhere to the collaboration of software and further exert the power and value of software, help customers and partners accelerate the creation of applications and push them to the market, and work with ecological partners to achieve greater innovation and breakthroughs in terms of tools, optimization, open source, talents, etc.
[ZOL client download] to see the latest technological information, search for "Zhongguancun Online" in the APP market, and the client reading experience is better. (7315612)
