Intel Vision 2024 Conference: Intel releases new hardware and software platform to help enterprises advance AI innovation at full speed

Released AI open system strategy and demonstrated collaboration with new customers and partners across all areas of AI.

• Intel has released a new AI strategy for enterprise customers. Its open and scalable features are widely applicable to all areas of AI.

• Announcing new branding for next-generation Intel® Xeon® 6 processors for data center, cloud, and edge.

• Intel® Gaudi 3 AI accelerator was launched, which has significantly improved reasoning capabilities and energy efficiency. Multiple OEM customers will adopt it, providing enterprises with a wider range of product choices in the AI ​​data center market.

• Intel, together with SAP, Red Hat, VMware and other industry leaders, announced that they will create an open platform to help enterprises drive AI innovation and accelerate the deployment of secure generative AI systems with the support of RAG (Retrieval Augmentation Generation) technology.

• Through the Ultra Ethernet Consortium (UEC), Intel is driving open Ethernet network innovation for AI high-speed interconnect technology (AI Fabrics). Intel has released a series of Ethernet solutions optimized for AI, including Intel AI Network Connection Cards (AI NICs) and AI connection chips.

On April 9, local time in the United States, Intel held the Intel on Industry Innovation Conference for customers and partners, announced the new brand of Intel® Xeon® 6 processors, launched the Intel Gaudi 3 accelerator, and helped companies promote generative AI innovation with high performance, openness and flexibility. It also released a full-stack solution covering new open and scalable systems, next-generation products and a series of strategic cooperation to accelerate the implementation of generative AI. According to the survey results of cnvrg.io, only 10% of companies successfully productized their generative AI projects in 2023. Intel's latest solution is expected to help companies cope with the challenges they face in promoting AI projects.

Pat Gelsinger, CEO of Intel, said: "Innovative technology is developing at an unprecedented speed, and every company is accelerating to become an AI company, all of which requires the support of semiconductor technology. From PCs to data centers to the edge, Intel is bringing AI into thousands of industries. Intel's latest Gaudi, Xeon and Core platforms will provide flexible and customizable solutions to meet the changing needs of customers and partners and seize the huge opportunities of the future."

Enterprises want to expand generative AI from the experimental stage to the application stage. Therefore, they need a rapidly deployable solution based on a high-performance, cost-effective, energy-efficient processor such as Intel Gaudi 3 AI accelerator. In addition, Gaudi 3 can also meet the needs of complexity, cost-effectiveness, fragmentation, data reliability and compliance.

Introducing Gaudi 3 for AI training and inference

Intel Gaudi 3 AI accelerators will power AI systems that can connect up to tens of thousands of accelerators via the universal standard of Ethernet. Compared with the previous generation, Intel Gaudi 3 will bring 4 times the BF16 AI computing power and 1.5 times the memory bandwidth. The accelerator will bring a major leap in AI training and reasoning for companies seeking to deploy generative AI at scale.

Intel Gaudi 3 is expected to significantly reduce the training time of the 7 billion and 13 billion parameter Llama2 models, as well as the 175 billion parameter GPT-3 model. In addition, it also demonstrated excellent performance in inference throughput and energy efficiency for the Llama 7B, 70B and Falcon 180B large language models (LLM).

Intel Gaudi 3 provides open, community-based software and industry-standard Ethernet networking, allowing enterprises to flexibly scale from a single node to clusters, super clusters, and ultra-large clusters with thousands of nodes, supporting large-scale reasoning, fine-tuning, and training. Intel Gaudi 3 will be available to OEMs in the second quarter of 2024.

Intel AI solutions create value for customers

Intel outlines its strategy for open, scalable AI systems, including hardware, software, frameworks, and tools. Intel enables a wide range of AI open ecosystem participants, such as device manufacturers, database providers, system integrators, software and service providers, to provide solutions that meet the specific generative AI needs of enterprises. At the same time, it also allows enterprises to collaborate with their known and trusted ecosystem partners and adopt corresponding solutions.

Ecosystem jointly creates an open platform to help enterprises innovate and develop AI

Intel, together with Anyscale, Articul8, DataStax, Domino, Hugging Face, KX Systems, MariaDB, MinIO, Qdrant, RedHat, Redis, SAP, VMware, Yellowbrick and Zilliz, announced that they will create an open platform to help enterprises drive AI innovation. This industry-wide plan aims to develop open, multi-vendor generative AI systems that provide first-class deployment convenience, performance and value through RAG (Retrieval Enhanced Generation) technology. RAG enables enterprises to enhance the large number of existing proprietary data sources running on standard cloud infrastructure with open large language model (LLM) capabilities, accelerating the application of generative AI in enterprises.

In the initial phase of the program, Intel will launch reference implementations for generative AI pipelines based on secure Xeon processors and Gaudi solutions, release technical conceptual frameworks, and continue to enhance the capabilities of the Intel Tiber Developer Cloud Platform infrastructure to lay the foundation for ecosystem development and validation of RAG and future processes. Intel encourages the ecosystem to further participate in the creation of this open platform to promote enterprise adoption of the platform, expand the scope of application of solutions, and achieve business results.

Intel Expands AI Roadmap and Open Ecosystem Approach

Intel also shared the latest information on next-generation products and services in each segment of enterprise AI.

New Intel® Xeon® 6 processors: Intel Xeon processors provide performance-efficient solutions for running current generative AI solutions, including RAGs that use proprietary data to generate specific business results. At the same time, Intel has also rebranded the next generation of processors for data centers, clouds, and edges, the Intel Xeon 6. Intel Xeon 6 processors with energy efficiency cores (E-cores) will be available in the second quarter of 2024, providing excellent efficiency, and Intel Xeon 6 processors with performance cores (P-cores) will follow shortly thereafter, bringing higher AI performance.

• Intel Xeon 6 processor with energy efficiency cores (codenamed Sierra Forest):

• Delivers 2.4x better performance per watt and 2.7x higher rack density than 2nd Generation Intel® Xeon® processors.

• Customers were able to replace old systems at a ratio of nearly 3:1, significantly reducing energy consumption and helping them achieve their sustainability goals.

  
  

• Intel Xeon 6 processor with performance cores (code-named Granite Rapids):

• Includes software support for the MXFP4 data format, which reduces latency to the next token by up to 6.5 times compared to 4th Generation Intel® Xeon® processors using FP16, and is capable of running the 70 billion parameter Llama-2 model.

Client, Edge and Connectivity: Intel shared momentum in client products and roadmap updates for a number of edge and networking products, including:

• Intel® Core™ Ultra processors provide new capabilities for productivity, security, and content creation, and provide tremendous power for enterprises to renew their PC devices. Intel expects to ship 40 million AI PCs by 2024, as well as more than 230 designs, covering thin and light PCs and game handheld devices.

  
  

• The next-generation Intel Core Ultra client processor family (codenamed Lunar Lake), which will be launched in 2024, will have a platform computing power of more than 100 TOPS and a computing power of more than 46 TOPS on the neural network processing unit (NPU), providing strong support for the next generation of AI PCs.

  
  

• Intel has released new edge chips covering Intel Core Ultra, Intel® Core™, Intel Atom® processors and Intel Resilience™ graphics series products, mainly targeting key areas such as retail, industrial manufacturing and medical care. All new products in Intel's edge AI product portfolio will be available this quarter and will be supported by Intel's Tiber edge platform this year.

  
  

• Through the Ultra Ethernet Consortium (UEC), Intel is driving open Ethernet network innovation for AI Fabrics and launching a series of Ethernet solutions optimized for AI. These innovations are designed to revolutionize AI high-speed interconnect technologies that can be massively scaled up and out to support the training and reasoning of AI models, which are becoming increasingly large and growing by an order of magnitude with each generation. Intel's product portfolio includes Intel AI Network Connection Cards (AI NICs), AI connection cores integrated into XPUs, Gaudi accelerator-based systems, and a series of AI interconnect hardware and software reference designs for Intel foundry.

Intel Tiber Business Solutions Portfolio

Intel released the Intel Tiber Business Solutions portfolio to simplify the deployment of enterprise software and services, including generative AI.

This unified experience will make it easier for enterprise customers and developers to find solutions that fit their needs, accelerate innovation and unlock value without sacrificing security, compliance or performance. Customers can start exploring the Intel Tiber portfolio today, and the solution will be generally available in the third quarter of this year.