Based on ecology, innovation casts soul Intel helps thousands of industries release AI potential with new software and hardware

AI cultural map in the second half of 2022 Software First, it ignited the popularity of generative AI, and then this craze reached its peak with the release of ChatGPT. Although the application scenario of generative AI was initially oriented towards entertainment creation, enterprises soon realized its role in improving their business competitiveness and tried to integrate generative AI into different business models. According to data, by 2026, 80% of enterprises will use generative AI, and 50% of enterprises will deploy various machine learning or deep learning applications in edge scenarios, thus improving enterprise competitiveness. The enterprise's spending on generative AI is expected to climb from $40 billion this year to $151 billion in 2027.

At present, AI is more responsible for some simple auxiliary functions, such as cultural map, outline refining, meeting summary, etc., and then with the maturity of related technologies, The era of AI assistant will come, when AI can help people predict habits and analyze data to take some simple actions, such as intelligent customer service, daily itinerary management assistant, or help enterprises answer compliance related questions. Then we will usher in the era of full-featured AI. AI assistants can perform complex tasks and functions through interaction, promote the automation of enterprise departments, and provide comprehensive and more accurate value based on data exchange of different AI.

However, both the era of AI assistant and the era of full function AI are based on the introduction of enterprise data and knowledge into the existing AI model. However, due to different data management methods, "enterprise" and "AI" are still two relatively independent concepts, because enterprise data is relatively more focused on privacy, but the general AI model is based on open data, In response, Zhuang Binghan, vice president of Intel's marketing group, general manager of data center sales in China and general manager of operator sales in China, said, "Intel has put forward the concept of enterprise AI in view of this situation, that is, through an open ecology, to achieve more large-scale and reliable help enterprises release AI potential, which is the most basic core of Intel's enterprise AI."

However, different levels also have different considerations. Enterprise AI needs more diversified underlying computing power, and needs accessibility and privacy; Secondly, AI infrastructure should be scalable and standardized, and support both private cloud and public cloud deployment; Then, the software needs to be safe and reliable, ideally based on open APIs or open source software; Finally, in terms of application ecology, it is also necessary to meet the needs of convenience and openness.

New Xeon 6 processor Provide a solid foundation for calculation

As the foundation of enterprise AI, the importance of computing power is beyond doubt. From the perspective of computing power, Intel can provide multiple heterogeneous computing power from cloud to edge to terminal. For enterprise AI, Intel Intel Vision 2024 introduces the next-generation Intel Xeon 6 processor for data center, cloud and edge.

Different from previous Xeon products, Xeon 6 is not a separate product, but a product portfolio, including Intel Xeon 6 processors (code name Sierra Forest) equipped with energy efficiency cores and Intel Xeon 6 performance core processors (code name Granite Rapids). The former has higher energy efficiency and is optimized for high-density and scale out workloads, The latter has higher performance, kernel density, memory and I/O innovation, as well as higher energy consumption. It is optimized for computing intensive and AI workloads.

In terms of architecture design, Sierra Forest and Granite Rapids all adopt a modular design method, which interconnects modular dies through fabric technology to achieve a more flexible architecture. In this way, independent computing and I/O chiplets can be combined more flexibly, and high bandwidth and low latency can be achieved with the help of EmiB packaging technology.

In terms of energy efficiency, compared with the second generation Intel Xeon processor, the performance per watt of Sierra Forest has been improved by 2.4 times, and the rack density has been increased by 2.7 times. This energy efficiency improvement not only significantly reduces the physical space occupied by equipment, but also saves more than 1 megawatt of jobs. Under the background of increasingly stringent requirements for energy conservation and emission reduction in China, Sierra Forest is undoubtedly an ideal choice.

In terms of performance, Granite Rapids can not only support MXFP4 data format, but also improve the performance of traditional FP16 format. It can even run Llama2 with 70 billion parameters, thus bringing higher AI performance to customers. In addition, according to the latest news, when Granite Rapids runs Llama 3 model reasoning with 8 billion parameters, the latency is twice that of the fourth generation Intel Xeon processors, and it can run Llama 3 model reasoning with 70 billion parameters at a token delay of less than 100 milliseconds.

Cooperate with ecological partners to help digital intelligence reform

Now, generative AI is restoring the core competitiveness of all enterprises, and has become an irreversible process. Correspondingly, the second half of AI must also be the stage from technological breakthrough to landing application. In Intel's view, the two keys to the implementation of AI strategy are "openness" and "easy deployment". "Intel has always adhered to the concept of" water conservancy without competition "in China, especially when the new technology trend is coming and the application of new generation technology is expanding, the open, reliable and easy to deploy strategy is extremely critical for end users, which is also the basis for our cooperation with the entire ecosystem, including OEM, ODM, ISV, SI, etc., in every previous technology iteration process. Technology is not born for technology, but for landing. " Liang Yali, vice president of marketing group of Intel Corporation and general manager of cloud and industry solutions department in China, said.

For many years, Intel has been working closely with China's ecological partners and promoting industrial development through a series of innovative product portfolios and solutions. For example, in the field of cloud computing, Intel has cooperated with Jinshan Cloud to introduce the fourth generation of Xeon scalable processors into its seventh generation performance guaranteed cloud server X7 for targeted optimization. At the same time, Intel has optimized three major major models, namely, Stable Diffusion, Llama2 and ChatGLM2, and released optimized model images. Compared with the model before optimization, the reasoning performance of the Stable Diffusion optimized by Intel AMX is improved by up to 4.96 times, while the performance of the open source big language models Llama2 and ChatGLM2 is improved by 2.62 times and 2.52 times.

Also in the field of cloud computing, Intel and JD Cloud have launched a number of landing attempts in AI scenarios. For example, in terms of intelligent marketing, JD has realized digital people who can provide services based on industry insight and cargo thinking, and the cost is extremely low; In terms of intelligent customer service, JD uses AI technology to achieve intelligent customer service with huge service volume and multimodal understanding. Behind these large model applications and cloud services is the cloud infrastructure based on Intel Xeon scalable processors. In terms of performance, compared with the server platform based on JD customized fourth generation Xeon scalable processors, the overall performance of JD Cloud's new generation cloud server based on the fifth generation Xeon scalable processors has increased by 23%.

In the field of intelligent manufacturing, TCL CStar has realized the improvement of production efficiency and the reduction of cost with the help of Intel. Specifically, TCL CStar has established the joint laboratory of TCL CStar Optoelectronics and Intel Intelligent Manufacturing through Intel's end-to-end technology capabilities, and has achieved the opening of closed-loop solutions such as device end drawing, end to end computing, terminal display, and result feedback. At the same time, in the production line inspection system and smart logistics, TCL Huaxing has reduced the cost of manpower and equipment configuration and significantly improved the product yield with the help of the hardware products and "cloud edge" architecture provided by Intel.

In the field of life science, traditional pharmaceutical processes often require a period of more than 10 years and huge R&D costs. The PandaOmics platform optimized by Intel and Intel can provide effective support for doctors in hospitals, researchers in scientific research institutions, and R&D personnel in pharmaceutical enterprises, With the support of Intel Core Ultra7 and the computing power based on the fifth generation Intel Xeon scalable processor, it can accelerate the whole set of drug research and development processes from target discovery, small molecule generation to clinical result prediction, and realize cost reduction and efficiency increase of pharmaceutical processes.

Finally, in the field of education, East China Normal University has developed the OpenEdu4ALL large model application platform with the help of Intel. It can not only significantly improve the efficiency and quality of teachers' lesson preparation, but also effectively improve teachers' digital literacy and realize intelligent education.

From theoretical exploration in the last century to the current generation AI boom, AI is reshaping life and society. As a leading semiconductor manufacturer, Intel has been committed to helping enterprises unleash data potential through open, scalable hardware and software platforms and end-to-end solutions for many years. Now, with the arrival of the AI era, Intel is also focusing on openness, easy deployment, scalability, and reliability. With the full stack of data center product portfolio, Intel is empowering thousands of businesses to accelerate innovation.

This article is an original article. If it is reproduced, please indicate the source: Based on ecology, innovation casts soul Intel helps thousands of industries release AI potential with new software and hardware https://server.zol.com.cn/867/8676076.html