At the Intel Vision 2024 customer and partner conference, Intel has unveiled a range of new open scalable systems, next-generation products, and strategic partnerships to speed up the adoption of enterprise generative artificial intelligence (GenAI). Intel also introduced the Intel Gaudi 3 accelerator to bring performance, openness, and choice to GenAI.
Intel’s newest products target the difficulties companies face when scaling AI efforts, because just 10% of firms were able to effectively implement GenAI experiments into production last year.
The Intel Gaudi 3 AI accelerator will power AI systems with up to tens of thousands of accelerators connected through the common standard of Ethernet. In comparison to its predecessor, the Intel Gaudi 3 would offer 1.5 times greater memory bandwidth and 4 times more AI compute for BF16. For multinational corporations wishing to use GenAI on a large scale, the accelerator would bring about a notable breakthrough in AI training and inference.
According to Intel itself, Intel Gaudi 3 is expected to provide 50 percent quicker time-to-train on average across Llama2 models with 7B and 13B parameters and GPT-3 175B parameter model when compared to Nvidia H100. Furthermore, it is anticipated that the inference throughput of the Intel Gaudi 3 accelerator would surpass that of the H100 by an average of 50 percent and 40 percent in the case of inference power-efficiency averaged2 across Falcon 180B parameter models and Llama 7B and 70B parameter models.
Features of Intel Gaudi 3 include open, community-based software and industry-standard Ethernet networking. Additionally, it would give businesses the flexibility to grow from a single node to hundreds of node clusters, super-clusters, and mega-clusters, providing inference, fine-tuning, and training at the biggest possible scale.
In the second quarter of 2024, OEMs including as Dell Technologies, HPE, Lenovo, and Supermicro will get access to Intel Gaudi 3.
“Innovation is advancing at an unprecedented pace, all enabled by silicon – and every company is quickly becoming an AI company,” said Intel CEO Pat Gelsinger. “Intel is bringing AI everywhere across the enterprise, from the PC to the data center to the edge. Our latest Gaudi, Xeon and Core Ultra platforms are delivering a cohesive set of flexible solutions tailored to meet the changing needs of our customers and partners and capitalize on the immense opportunities ahead.”
Intel’s Expanded AI Roadmap with Intel Xeon 6
New Intel Xeon Processors – New Intel Xeon processors would provide effective ways to operate existing GenAI systems, such as RAG, which use private data to provide outcomes tailored to a business’s needs. Intel unveiled the Intel Xeon 6, the new name for its next-generation CPUs for cloud, edge, and data centers. This quarter, Intel Xeon 6 processors with new Efficient-cores (E-cores) will go on sale. Meanwhile, Intel Xeon 6 processors with Performance-cores (P-cores) will go on sale shortly after the E-core processors and will provide better AI performance.
Intel Xeon 6 processors with E-cores (code-named Sierra Forest) would come with the following specs, according to Intel:
- Comparing with 2nd Gen Intel Xeon processors, there is a 4x increase in performance per watt and a 2.7x boost in rack density
- Older systems may be replaced by customers almost three to one, significantly reducing energy use and assisting in the achievement of sustainability objectives
Intel Xeon 6 processors with P-cores (code-named Granite Rapids) would come with the following specs, according to Intel:
- Include software support for the MXFP4 data format, which can run 70 billion parameter Llama-2 models and minimize next token latency by up to 6.5x when compared to 4th Gen Intel Xeon processors utilizing FP16
Creating an Open Platform for Business AI
At the Intel Vision 2024 conference, Intel also announced plans to develop an open platform for corporate AI in partnership with Anyscale, Articul8, DataStax, Domino, Hugging Face, KX Systems, MariaDB, MinIO, Qdrant, RedHat, Redis, SAP, VMware, Yellowbrick, and Zilliz. With the help of retrieval-augmented generation, the industry-wide project seeks to create open, multivendor GenAI systems that provide the highest performance, value, and simplicity of implementation available. RAG accelerates the implementation of GenAI in organizations by enabling the addition of open LLM capabilities to the massive, current proprietary data sources that are operating on standard cloud infrastructure.
To facilitate ecosystem development and validation of RAG and future pipelines, Intel will first release reference implementations for GenAI pipelines on secure Intel Xeon and Gaudi-based solutions, publish a technical conceptual framework, and keep adding infrastructure capacity to the Intel Tiber Developer Cloud. Intel invites more ecosystem members to collaborate in this open endeavor to expedite business outcomes, expand solution coverage, and ease corporate adoption.
Intel Tiber Portfolio of Business Solutions
The Intel Tiber portfolio of business solutions, which includes GenAI, was also introduced by Intel during the Intel Vision 2024 conference with the goal of streamlining the adoption of corporate software and services.
Without sacrificing security, compliance, or performance, a single experience would make it simpler for business customers and developers to choose solutions that meet their requirements, spur creativity, and generate value. The Intel Tiber portfolio is now available for customers to explore; the complete deployment is scheduled for the third quarter of 2024.
Targeting key sectors including retail, industrial production, and healthcare, Intel unveiled new edge silicon for the Intel Core Ultra, Intel Core, and Intel Atom CPU families as well as the Intel Arc graphics processing unit (GPU) family of devices. This quarter, all new products that Intel has added to its edge AI portfolio will be available. The Intel Tiber Edge Platform will support these solutions as well.
