Intel unveiled cutting-edge technologies and architectures that will dramatically accelerate the AI ecosystem from data centers, clouds, and networks to the edge and PCs at Computex in Taiwan on the 4th.

Key contents include: △Launch of the Intel Xeon 6 E-Core Processor △Pricing policy for the Intel Gaudi 2 and Intel Gaudi 3 AI accelerator kits △Luna Lake client processor architecture that will continue the growth of AI PCs, etc.

“AI is driving the most significant era of innovation in the history of our industry,” said Intel CEO Pat Gelsinger. “Semiconductors are once again enabling exponential advances in computing performance that will push the boundaries of human potential and drive the global economy for years to come.”

“Intel’s latest Xeon, Gaudi and Core Ultra platforms, combined with the capabilities of Intel’s hardware and software ecosystem, provide customers with flexible, secure, sustainable and cost-effective solutions,” said Pat Gelsinger, CEO.

■ Intel, AI Everywhere Implementation
In his Computex keynote, Gelsinger emphasized open standards and Intel’s strong ecosystem that will accelerate AI opportunities. Industry players including Acer, ASUS, Microsoft, and Inventec joined in to express their support.

Gelsinger CEO and industry leaders made it clear that Intel is leading the way in AI innovation and delivering next-generation technologies ahead of schedule. Intel introduced the first Xeon 6 products following the launch of the 5th generation Intel Xeon processors in just six months.

We previewed the Gaudi AI accelerator and provided a cost-effective, high-performance generative AI training and inference system to enterprise customers. Additionally, it opened the era of AI PCs by equipping more than 8 million devices with Intel Core Ultra processors, and also unveiled the client architecture scheduled for release later this year.

Through these advancements, Intel is democratizing AI and enabling the industry by accelerating execution speed while pushing the boundaries of innovation and production speed.

■ Intel Xeon 6 Processor Unveiled
As digital transformation accelerates, enterprises face pressure to replace aging data center systems to reduce costs, meet sustainability goals, maximize physical and rack space utilization, and create new digital capabilities across the enterprise.

The Xeon 6 platform and processor family is designed to address these challenges with E-core (Efficient-core) and P-core (Performance-core) models.

Addressing a wide range of workloads and use cases, from AI and other high-performance computing requirements to scalable cloud-native applications, both E-Core and P-Core are built on an architecture that is compatible with a common software stack and an open ecosystem of hardware and software vendors.

The Xeon 6 E-Core delivers performance and power efficiency for high-density scale-out workloads in data centers, featuring 3:1 rack consolidation, up to a 4.2x increase in rack-level performance, and up to a 2.6x increase in performance-per-watt.

Intel Xeon 6 E-Core is the best It was first released and is said to be available now. The Xeon 6 P-core, codenamed Granite Rapids, is expected to be released next quarter.

■ Providing high-performance generative AI with Intel Gaudi AI accelerator

Today, leveraging the power of generative AI has become faster and cheaper. The core infrastructure, x86, operates at scale in virtually every data center environment, providing cost-effective interoperability and the benefits of an open ecosystem of developers and customers, while serving as the foundation for integrating AI capabilities.

The Intel Gaudi 3 accelerator supports performance for generative model training and inference workloads. With an 8,192-accelerator cluster, Intel Gaudi 3 is expected to deliver up to 40 percent faster training times than a similar-sized NVIDIA H100 GPU cluster, and up to 15 percent faster training throughput for a 70 billion Llama2 (Llama2-70B) model on the NVIDIA H100 for a 64-accelerator cluster.

Additionally, Intel said that Gaudi 3 will provide up to 2x faster inference on average compared to NVIDIA H100 when running LLMs such as Llama2-70B and Mistral-7B.

Intel is working with global system suppliers to deliver AI systems. New partners joining at Computex this year include ASUS, Foxconn, Gigabyte, Inventec, Quanta, and Wistron, and product supply is expanding through major system suppliers such as Dell, HPE, Lenovo, and Supermicro.

■ On-device AI acceleration for notebook PCs
The AI PC space is changing every aspect of the computing experience, and Intel is at the forefront of creating this new category. It’s no longer just about faster processing speeds or sleeker designs, but also creating edge devices that anticipate customer needs and adapt to their preferences in real time, ushering in a whole new era of productivity, efficiency, and creativity.

With AI PCs expected to account for 60% of new PCs by 2027, Intel has been quick to build the best hardware and software platform for AI PCs. Working with more than 100 independent software vendors (ISVs), we are delivering 300 features and supporting 500 AI models across the Core Ultra platform.

On this day, Intel disclosed details about the architecture of Luna Lake, its next-generation flagship processor for AI PCs. It focuses on advanced performance in graphics and AI processing based on x86 and power-efficient computing performance for thin and light designs, and emphasizes that it provides up to 40% SoC power reduction and more than 3 times AI computing compared to the previous generation.

The 4th generation Intel NPU, which delivers up to 48 TOPS and 40 TOPS of AI performance, delivers up to 4x the AI compute of its predecessor, improving generative AI. The new GPU design, codenamed Battlemage, combines new innovations in two parts: the Xe2 GPU cores for graphics and the Xe Matrix Extension (XMX) array for AI.

The Xe2 GPU cores deliver 1.5x faster gaming and graphics performance than the previous generation, and the new XMX array supports a second AI accelerator with up to 67 TOPS of performance for exceptional throughput in AI content creation.

Luna Lake will be available in over 80 different AI PC designs from 20 PC manufacturers, and Intel expects to ship over 40 million core Ultra processors this year. It is expected to hit the market in the third quarter of 2024, targeting the holiday season.