Intel Just Built a 256-Core Xeon

Intel has officially revealed one of its most ambitious server processors yet: the next-generation Xeon platform codenamed Diamond Rapids. The flagship configuration will scale to up to 256 CPU cores, putting Intel directly into the same maximum core-count class as AMD’s upcoming EPYC 9006 “Venice” processors. And these are not efficiency-focused E-cores. Diamond Rapids is Intel’s next-generation P-core Xeon architecture, designed for high-performance enterprise computing, cloud infrastructure, HPC and increasingly demanding AI workloads. Intel says Diamond Rapids will be manufactured using its enhanced Intel 18A-P process and will use advanced packaging technologies including Foveros Direct 3D and UCIe-S interconnects. The specifications are enormous. Intel has confirmed up to 256 cores, as much as 1.28 GB of last-level cache, 16 memory channels supporting speeds up to 12,800 MT/s, and 128 lanes of PCIe Gen 6 with CXL 3.0 support. That represents a major jump from Granite Rapids Xeon processors, whose P-core configurations topped out at 128 cores. Memory bandwidth is another major focus. Diamond Rapids doubles down on a 16-channel memory subsystem designed to feed hundreds of CPU cores and AI workloads without turning memory access into a major bottleneck. Intel is also adding new Advanced Performance Extensions and enhanced Advanced Matrix Extensions to accelerate compute-intensive and AI-oriented workloads. The obvious competitor is AMD. AMD’s sixth-generation EPYC 9006 processors, codenamed Venice, use the Zen 6 architecture and scale to 256 cores and 512 threads per socket. They also support 16 DDR5 memory channels, PCIe 6.0 and high-speed MRDIMM memory. So for the first time, Intel and AMD will both compete at the 256-core level in high-end server CPUs. But raw core count will not decide the winner. Performance per core, memory bandwidth, power consumption, software optimization, price and total platform efficiency will all matter. Intel is positioning Diamond Rapids as a foundation for enterprise-scale agentic AI, where CPUs increasingly work alongside GPUs and dedicated accelerators rather than simply running traditional server applications. Diamond Rapids is currently scheduled to arrive in 2027, with the next-generation Coral Rapids architecture planned to follow afterward. The server CPU race is no longer about dozens of cores. It is now moving into the hundreds. Can Intel’s 256-core Xeon challenge AMD EPYC Venice? Subscribe to Xenon Lab News for more AI hardware, processors and semiconductor technology. #Intel #Xeon #DiamondRapids #Xeon7 #AMD #EPYC #Zen6 #ServerCPU #256Cores #AIHardware #DataCenter #PCIe6 #Intel18A #HPC #CloudComputing #Semiconductors #TechNews #XenonLabNews