In a strategic move to regain dominance in the data center market, Intel has unveiled its latest lineup of data center processors, directly challenging AMD’s advancements. The new Intel Xeon processors and AI accelerators are designed to cater to the growing demands of artificial intelligence and data-intensive applications.
The New Xeon Processors
Intel’s 5th-generation Xeon Scalable processors, codenamed “Emerald Rapids,” boast significant performance improvements over their predecessors. According to Intel, these processors offer a 21% average compute performance gain and 16% better memory speeds compared to the previous generation, “Sapphire Rapids”. The new chips support up to 64 cores with an L3 cache of 320MB, which is more than double the cache size of the earlier models. This increased cache size and simplified chiplet architecture contribute to a 1.21× performance gain.
AI and Sustainability Focus
Intel has also integrated AI acceleration into the Xeon processors, making them capable of handling large AI models without additional accelerators. The new Xeon processors deliver up to 42% higher inference and fine-tuning on models as large as 20 billion parameters. Additionally, these processors are touted as Intel’s most sustainable data center processors to date, offering 36% higher average performance per watt across various workloads.
Gaudi3 AI Accelerator
Alongside the Xeon processors, Intel introduced the Gaudi3 AI accelerator, developed by Habana Labs, a company Intel acquired in 2019. The Gaudi3 is designed to compete with Nvidia’s H100 and AMD’s MI300X GPUs, promising four times higher BF16 performance, two times faster networking performance, and one and a half times higher bandwidth than its predecessor, Gaudi2. This AI accelerator is aimed at enhancing AI training capabilities, a crucial aspect for modern data centers.
Competitive Landscape
The release of these new processors is a part of Intel’s broader strategy to fend off increasing competition from AMD and Nvidia in the data center market. AMD has made significant strides with its EPYC processors, which have been well-received for their performance and efficiency. Intel’s new offerings aim to close the performance gap and provide customers with more robust options for AI and data-intensive applications.
Intel’s data center roadmap includes the upcoming Sierra Forest and Granite Rapids architectures. Sierra Forest, expected in the first half of 2024, will feature up to 144 cores per chip, surpassing AMD’s current offerings. Granite Rapids will follow shortly after, further enhancing Intel’s data center capabilities.
Intel’s new data center chips represent a significant step in the company’s efforts to regain market share and address the growing demands of AI and data processing workloads. By improving performance, efficiency, and sustainability, Intel aims to provide a competitive edge in the ever-evolving data center landscape.
