Intel updated its data sensor processor roadmap at Investor Day.

Intel held an investor day on the 30th and shared the roadmap and progress of its data center business and AI group.

On this day, Intel unveiled updates to its next-generation Intel® Xeon® processors, updates to its 4th-generation Intel® Xeon® Scalable processors (codenamed Sapphire Rapids), and Intel’s hardware performance for a range of workloads, including competitive and AI-intensive workloads.

First of all, the ramp-up to mass production of the 4th generation Intel Xeon Scalable Processor (Sapphire Rapids) is progressing smoothly.

Currently, all major OEMs and ODMs are shipping systems based on 4th Gen Xeon systems, and the top 10 global CSPs are working to deploy services based on 4th Gen Xeon.

The 5th generation Intel Xeon Scalable processors (Emerald Rapids) are scheduled to be produced in the fourth quarter of 2023.

Emerald Rapids is Intel's Performance Core (P-Core) product. On the 30th, Intel officially named it the 5th generation Intel® Xeon® Scalable processor.

Intel is providing samples of the product to some customers.

Emerald Rapids uses the Intel 7 process, and as it is the successor to Sapphire Rapids, there are no expected delays in development or mass production.

Intel's first Efficient Core (E-Core) Xeon processor, codenamed Sierra Forest, is on track for production launch in the first half of 2024.

Intel revealed that Sierra Forest will feature 144 cores per socket.

Sierra Forest is also the first CPU based on Intel's upcoming 3rd generation process, which set a new record earlier this quarter by running multiple operating systems in less than a day.

Intel® Xeon® processors, codenamed Granite Rapids, are expected to be introduced following the Sierra Forest launch in 2024.

Although the specific schedule has not been revealed yet, it is expected that the release production period will be shortened as it shares a platform with Sierra Forest.

“We are building the world’s fastest memory interface for Granite Rapids,” said VP Spellman, while running a command prompt to demonstrate the memory configuration of Granite Rapids. “Intel is leading the ecosystem to develop and support a new type of DIMM called Multiplexer Composite Rank (MCR) based on DDR5 that can achieve transfer speeds of 8,800 MT/s.”

Clearwater Forest is scheduled for production in 2025 and will be built on Intel's 18A process.

The reason Intel is increasing the number of processor models is to provide processors optimized for various workloads.To provide this to our customers.

We are currently securing a diverse product portfolio as various types of computing emerge, such as the expansion of demand for AI operations.

Intel forecasts that the annual growth rate of data center processors will be in the low 20% range over the next five years, an upgrade from its previous forecast of mid-teens%.

Intel simplifies its GPU roadmap to allow for choice and focus.

In addition, it announced that it would separate the cost structure of its internal foundry and process design divisions and operate the system in a way that its own design CPUs are also distributed to the foundry.

High-load workloads such as AI, networking, and HPC are currently the fastest-growing workloads, accounting for 15% of the total. Habana Labs is preparing to launch the Gaudi 3 accelerator on TSMC’s 5nm process.

The market size of machine learning-related processors is expected to reach $40 billion by 2027.

It is expected that 60% of the market will be general CPUs, and 40% will be accelerators.

In addition to Intel’s efforts to foster an open and ethical ecosystem, Intel Xeon Scalable processors are also playing a role in generative AI.

Sandra Rivera, general manager, told investors that NVIDIA is using fourth-generation Xeons with H100 GPUs as head nodes to run virtual machines that accelerate generative AI models on Microsoft Azure, including ChatGPT.