In today’s world where artificial intelligence and cloud computing technologies are growing rapidly, the process power requirement of data centers is increasing every second. Intel, one of the pioneers of technological transformation, officially announces the Intel Xeon 6+ (Clearwater Forest) series, the new generation processor family developed for information centers and artificial intelligence infrastructures.

Based on the Intel 18A node, the company’s most advanced production architecture, this new family offers disruptive performance, especially in high process density, top-level power efficiency and cloud-based workloads. Intel Xeon 6990E+, the flagship of the series, aims to radically change the standards in the sector with its 288 efficiency cores (E-core) and huge cache capacity.

• First Fruit of Intel 18A Architecture:The Xeon 6+ series goes down in history as the first processor family in which Intel uses the most advanced 18 angstrom (Intel 18A) production technology in the information center segment.

• 288 Cores and 576 MB L3 Cache:The flagship model, the Xeon 6990E+, features 288 next-generation Darkmont E-cores and can scale up to 576 cores in total in dual-socket systems.

• Focused on Artificial Intelligence and Network Infrastructure:The processors provide up to 30 percent performance advantage per thread compared to competing architectures in spy-based artificial intelligence models, information analytics and cloud-native workloads.

What Does Intel 18A and Advanced Packaging Technology Offer?

When designing the Xeon 6+ architecture, Intel not only increases the number of cores, but also completely renews the basic structure of the processor. The Intel 18A manufacturing process used in the processor combines the RibbonFET transistor architecture and the PowerVia system that provides power distribution from the backside of the processor chip.

This innovative approach maximizes performance per watt by optimizing electron flow.

A disaggregate, that is, chiplet-based approach is adopted in the physical design of the processor. Thanks to EMIB 2.5D and Foveros Direct 3D packaging technologies, components from different production nodes are brought together flawlessly on a single silicon.

The Xeon 6+ platform includes 12 CPU chiplets produced with the Intel 18A process, and each chiplet contains 24 Darkmont cores that do not support Hyper-Threading. These massive processor cores rise above three base layers produced with Intel 3 architecture. On the outermost part, there are two Intel 7-based I/O (input-output) blocks that manage the data flow.

Bottlenecks in Memory and I/O Infrastructure Become History

One of the biggest issues when running large information sets and complex artificial intelligence algorithms is memory bottlenecks. Intel is making significant investments in the memory subsystem to prevent this problem. The Xeon 6+ series comes with a full 12-channel DDR5 8000 MT/s memory support. This high-speed memory bandwidth increases the speed at which servers process information.

PCIe Gen 5 and Hardware Accelerators

The platform offers 96 lanes of PCIe Gen 5 support for high-speed data transfer. In addition, thanks to the 64-line CXL (Compute Express Link) technology, direct, very low-latency contacts can be established with external accelerators and memory pools.

Eight integrated hardware accelerators located within the I/O blocks; It takes specific tasks such as cryptography, information flow processing (DSA) and in-memory analysis (IAA) directly from the CPU, allowing the main processing units to focus only on tasks that require pure performance.

New Era in Energy Tracking: Intel AET

Power consumption plays a critical role in the operating costs of data centers. One of the most striking innovations offered in Xeon 6+ models is Intel Application Energy Telemetry (AET) technology.

This hardware-based telemetry tool allows server owners to instantly monitor power consumption at the level of microservices, virtual machines, containers, and even personal software threads. In this way, cloud providers can perform high-level power optimization by precisely analyzing which application consumes how much power.

Power Consumption and Release Date

Intel states that the Xeon 6+ processor family is fully compatible with the LGA 4710 socket structure used in existing Xeon 6 platforms and can be easily integrated into existing infrastructures. Thermal design powers (TDP) of the processors vary between 300 W and 500 W, in parallel with the tremendous performance offered.

Claiming an average of 30 percent performance and 30 percent power efficiency advantage per thread against AMD’s 192-core Epyc 9965 model, the new generation monsters open the door to a new era focused on efficiency in data centers.