At Hot Chips 2026, Intel outlined the architecture of Diamond Rapids. The next-generation Xeon, built on the 18A-P process, will feature up to 256 cores, 1.28 GB of last-level cache, 16 memory channels running at 12,800 MT/s, and 128 PCIe Gen6 lanes. Intel is targeting the chip at agent-based AI in enterprise environments.
Intel explains that Diamond Rapids is built entirely on its own technology, including the 18A-P process, which promises improvements in power consumption and performance. The chip introduces a new architecture that combines adaptable compute building blocks, a unified memory fabric, and flexible I/O.
The package design received significant attention at Hot Chips. Intel is adopting Foveros Direct 3D and the UCIe-S interconnect, thereby moving away from EMIB. Intel also mentioned a high-bandwidth memory subsystem, the new Advanced Performance Extensions (APX), and expanded Advanced Matrix Extensions (AMX). These AMX instructions are designed to accelerate AI workloads directly on the CPU.
Crescent Island and Wildcat Lake
In addition to Diamond Rapids, Intel also provided an update on the Crescent Island inference GPU and the Wildcat Lake SoC for laptops and edge computing.
Crescent Island combines 32 Xe cores and 256 XMX engines based on Xe3P with up to 480 GB of LPDDR5X, packaged in a 350-watt air-cooled PCIe card. Initial customer shipments are scheduled for the second half of 2026.
The third chip is already on the market as the Intel Core Series 3. Wildcat Lake features two performance cores and four efficiency cores, Xe3 graphics with XMX acceleration, and an NPU delivering up to 17 TOPS. Wildcat Lake is also the first Intel processor to utilize UCIe for multi-chip packages.
Tip: Intel capitalizes on stock rally for billion-dollar offering